|
a |
|
b/datasets_csv/.ipynb_checkpoints/Preprocessing-checkpoint.ipynb |
|
|
1 |
{ |
|
|
2 |
"cells": [ |
|
|
3 |
{ |
|
|
4 |
"cell_type": "code", |
|
|
5 |
"execution_count": 1, |
|
|
6 |
"metadata": {}, |
|
|
7 |
"outputs": [], |
|
|
8 |
"source": [ |
|
|
9 |
"import os\n", |
|
|
10 |
"from os.path import join\n", |
|
|
11 |
"\n", |
|
|
12 |
"import pandas as pd\n", |
|
|
13 |
"import numpy as np\n", |
|
|
14 |
"\n", |
|
|
15 |
"label_col = 'survival_months'\n", |
|
|
16 |
"n_bins = 4\n", |
|
|
17 |
"eps = 1e-6" |
|
|
18 |
] |
|
|
19 |
}, |
|
|
20 |
{ |
|
|
21 |
"cell_type": "code", |
|
|
22 |
"execution_count": 2, |
|
|
23 |
"metadata": {}, |
|
|
24 |
"outputs": [], |
|
|
25 |
"source": [ |
|
|
26 |
"def add_bins(slide_data):\n", |
|
|
27 |
" assert 'case_id' in slide_data.columns and 'censorship' in slide_data.columns\n", |
|
|
28 |
" \n", |
|
|
29 |
" patients_df = slide_data.drop_duplicates(['case_id']).copy()\n", |
|
|
30 |
" uncensored_df = patients_df[patients_df['censorship'] < 1]\n", |
|
|
31 |
" disc_labels, q_bins = pd.qcut(uncensored_df[label_col], q=n_bins, retbins=True, labels=False)\n", |
|
|
32 |
" q_bins[-1] = slide_data[label_col].max() + eps\n", |
|
|
33 |
" q_bins[0] = slide_data[label_col].min() - eps\n", |
|
|
34 |
"\n", |
|
|
35 |
" disc_labels, q_bins = pd.cut(patients_df[label_col], bins=q_bins, retbins=True, labels=False, right=False, include_lowest=True)\n", |
|
|
36 |
" patients_df.insert(2, 'label', disc_labels.values.astype(int))\n", |
|
|
37 |
"\n", |
|
|
38 |
" patient_dict = {}\n", |
|
|
39 |
" slide_data = slide_data.set_index('case_id')\n", |
|
|
40 |
" for patient in patients_df['case_id']:\n", |
|
|
41 |
" slide_ids = slide_data.loc[patient, 'slide_id']\n", |
|
|
42 |
" if isinstance(slide_ids, str):\n", |
|
|
43 |
" slide_ids = np.array(slide_ids).reshape(-1)\n", |
|
|
44 |
" else:\n", |
|
|
45 |
" slide_ids = slide_ids.values\n", |
|
|
46 |
" patient_dict.update({patient:slide_ids})\n", |
|
|
47 |
" \n", |
|
|
48 |
" return q_bins, patient_dict, patients_df" |
|
|
49 |
] |
|
|
50 |
}, |
|
|
51 |
{ |
|
|
52 |
"cell_type": "code", |
|
|
53 |
"execution_count": 3, |
|
|
54 |
"metadata": {}, |
|
|
55 |
"outputs": [], |
|
|
56 |
"source": [ |
|
|
57 |
"slide_data = pd.read_csv('./tcga_gbmlgg_all_clean.csv.zip', compression='zip', header=0, index_col=0, sep=',', low_memory=False)\n", |
|
|
58 |
"\n", |
|
|
59 |
"n_bins = 4\n", |
|
|
60 |
"eps = 1e-6\n", |
|
|
61 |
"\n", |
|
|
62 |
"### Asserts that 'case_id' is a column, not an index.\n", |
|
|
63 |
"if 'case_id' not in slide_data:\n", |
|
|
64 |
" slide_data.index = slide_data.index.str[:12]\n", |
|
|
65 |
" slide_data['case_id'] = slide_data.index\n", |
|
|
66 |
" slide_data = slide_data.reset_index(drop=True)\n", |
|
|
67 |
"\n", |
|
|
68 |
"q_bins, patients_dict, slide_data = add_bins(slide_data)\n", |
|
|
69 |
"\n", |
|
|
70 |
"slide_data.reset_index(drop=True, inplace=True)\n", |
|
|
71 |
"slide_data = slide_data.assign(slide_id=slide_data['case_id'])\n", |
|
|
72 |
"\n", |
|
|
73 |
"label_dict = {}\n", |
|
|
74 |
"key_count = 0\n", |
|
|
75 |
"for i in range(len(q_bins)-1):\n", |
|
|
76 |
" for c in [0, 1]:\n", |
|
|
77 |
" label_dict.update({(i, c):key_count})\n", |
|
|
78 |
" key_count+=1\n", |
|
|
79 |
"\n", |
|
|
80 |
"for i in slide_data.index:\n", |
|
|
81 |
" key = slide_data.loc[i, 'label']\n", |
|
|
82 |
" slide_data.at[i, 'disc_label'] = key\n", |
|
|
83 |
" censorship = slide_data.loc[i, 'censorship']\n", |
|
|
84 |
" key = (key, int(censorship))\n", |
|
|
85 |
" slide_data.at[i, 'label'] = label_dict[key]\n", |
|
|
86 |
"\n", |
|
|
87 |
"bins = q_bins\n", |
|
|
88 |
"num_classes=len(label_dict)\n", |
|
|
89 |
"patients_df = slide_data.drop_duplicates(['case_id'])\n", |
|
|
90 |
"patient_data = {'case_id':patients_df['case_id'].values, 'label':patients_df['label'].values}\n", |
|
|
91 |
"\n", |
|
|
92 |
"new_cols = list(slide_data.columns[-2:]) + list(slide_data.columns[:-2])\n", |
|
|
93 |
"slide_data = slide_data[new_cols]\n", |
|
|
94 |
"metadata = slide_data.columns[:11]" |
|
|
95 |
] |
|
|
96 |
}, |
|
|
97 |
{ |
|
|
98 |
"cell_type": "code", |
|
|
99 |
"execution_count": 5, |
|
|
100 |
"metadata": {}, |
|
|
101 |
"outputs": [], |
|
|
102 |
"source": [ |
|
|
103 |
"from sklearn.pipeline import Pipeline\n", |
|
|
104 |
"from sklearn.decomposition import PCA\n", |
|
|
105 |
"from sklearn.preprocessing import StandardScaler\n", |
|
|
106 |
"\n", |
|
|
107 |
"\n", |
|
|
108 |
"def series_intersection(s1, s2):\n", |
|
|
109 |
" return pd.Series(list(set(s1) & set(s2)))\n", |
|
|
110 |
"\n", |
|
|
111 |
"genomic_features = slide_data.drop(metadata, axis=1)\n", |
|
|
112 |
"scaler_omic = StandardScaler().fit(genomic_features)" |
|
|
113 |
] |
|
|
114 |
}, |
|
|
115 |
{ |
|
|
116 |
"cell_type": "code", |
|
|
117 |
"execution_count": 9, |
|
|
118 |
"metadata": {}, |
|
|
119 |
"outputs": [ |
|
|
120 |
{ |
|
|
121 |
"name": "stderr", |
|
|
122 |
"output_type": "stream", |
|
|
123 |
"text": [ |
|
|
124 |
"/home/mahmoodlab/anaconda3/lib/python3.8/site-packages/IPython/core/interactiveshell.py:3071: DtypeWarning: Columns (2) have mixed types.Specify dtype option on import or set low_memory=False.\n", |
|
|
125 |
" has_raised = await self.run_ast_nodes(code_ast.body, cell_name,\n" |
|
|
126 |
] |
|
|
127 |
} |
|
|
128 |
], |
|
|
129 |
"source": [ |
|
|
130 |
"signatures = pd.read_csv('./signatures.csv')\n", |
|
|
131 |
"slide_df = pd.read_csv('./tcga_gbmlgg_all_clean.csv.zip')" |
|
|
132 |
] |
|
|
133 |
}, |
|
|
134 |
{ |
|
|
135 |
"cell_type": "code", |
|
|
136 |
"execution_count": 40, |
|
|
137 |
"metadata": {}, |
|
|
138 |
"outputs": [], |
|
|
139 |
"source": [ |
|
|
140 |
"omic_from_signatures = []\n", |
|
|
141 |
"for col in signatures.columns:\n", |
|
|
142 |
" omic = signatures[col].dropna().unique()\n", |
|
|
143 |
" omic_from_signatures.append(omic)\n", |
|
|
144 |
"\n", |
|
|
145 |
"omic_from_signatures = np.concatenate(omic_from_signatures)" |
|
|
146 |
] |
|
|
147 |
}, |
|
|
148 |
{ |
|
|
149 |
"cell_type": "code", |
|
|
150 |
"execution_count": 41, |
|
|
151 |
"metadata": {}, |
|
|
152 |
"outputs": [ |
|
|
153 |
{ |
|
|
154 |
"name": "stderr", |
|
|
155 |
"output_type": "stream", |
|
|
156 |
"text": [ |
|
|
157 |
"/home/mahmoodlab/anaconda3/lib/python3.8/site-packages/IPython/core/interactiveshell.py:3071: DtypeWarning: Columns (2) have mixed types.Specify dtype option on import or set low_memory=False.\n", |
|
|
158 |
" has_raised = await self.run_ast_nodes(code_ast.body, cell_name,\n", |
|
|
159 |
"/home/mahmoodlab/anaconda3/lib/python3.8/site-packages/IPython/core/interactiveshell.py:3071: DtypeWarning: Columns (4) have mixed types.Specify dtype option on import or set low_memory=False.\n", |
|
|
160 |
" has_raised = await self.run_ast_nodes(code_ast.body, cell_name,\n" |
|
|
161 |
] |
|
|
162 |
} |
|
|
163 |
], |
|
|
164 |
"source": [ |
|
|
165 |
"for fname in os.listdir('./'):\n", |
|
|
166 |
" if fname.endswith('.csv.zip'):\n", |
|
|
167 |
" slide_df = pd.read_csv(fname)\n", |
|
|
168 |
" omic_overlap = np.concatenate([omic_from_signatures+mode for mode in ['_mut', '_cnv', '_rnaseq']])\n", |
|
|
169 |
" omic_overlap = sorted(series_intersection(omic_overlap, slide_df.columns))\n", |
|
|
170 |
" slide_df[list(slide_df.columns[:9]) + omic_overlap].to_csv('../dataset_csv/%s' % fname)" |
|
|
171 |
] |
|
|
172 |
}, |
|
|
173 |
{ |
|
|
174 |
"cell_type": "code", |
|
|
175 |
"execution_count": 36, |
|
|
176 |
"metadata": {}, |
|
|
177 |
"outputs": [], |
|
|
178 |
"source": [ |
|
|
179 |
"omic_from_signatures = []\n", |
|
|
180 |
"for col in signatures.columns:\n", |
|
|
181 |
" omic = signatures[col].dropna().unique()\n", |
|
|
182 |
" omic_from_signatures.append(omic)\n", |
|
|
183 |
"\n", |
|
|
184 |
"omic_from_signatures = np.concatenate(omic_from_signatures)\n" |
|
|
185 |
] |
|
|
186 |
}, |
|
|
187 |
{ |
|
|
188 |
"cell_type": "code", |
|
|
189 |
"execution_count": 7, |
|
|
190 |
"metadata": {}, |
|
|
191 |
"outputs": [ |
|
|
192 |
{ |
|
|
193 |
"name": "stdout", |
|
|
194 |
"output_type": "stream", |
|
|
195 |
"text": [ |
|
|
196 |
"Tumor Suppressor Genes Embedding Size: 84\n", |
|
|
197 |
"Oncogenes Embedding Size: 314\n", |
|
|
198 |
"Protein Kinases Embedding Size: 498\n", |
|
|
199 |
"Cell Differentiation Markers Embedding Size: 415\n", |
|
|
200 |
"Transcription Factors Embedding Size: 1396\n", |
|
|
201 |
"Cytokines and Growth Factors Embedding Size: 428\n" |
|
|
202 |
] |
|
|
203 |
} |
|
|
204 |
], |
|
|
205 |
"source": [ |
|
|
206 |
"\n", |
|
|
207 |
"def series_intersection(s1, s2):\n", |
|
|
208 |
" return pd.Series(list(set(s1) & set(s2)))\n", |
|
|
209 |
"\n", |
|
|
210 |
"sig_names = []\n", |
|
|
211 |
"for col in signatures.columns:\n", |
|
|
212 |
" sig = signatures[col].dropna().unique()\n", |
|
|
213 |
" sig = np.concatenate([sig+mode for mode in ['_mut', '_cnv', '_rnaseq']])\n", |
|
|
214 |
" sig = sorted(series_intersection(sig, genomic_features.columns))\n", |
|
|
215 |
" sig_names.append(sig)\n", |
|
|
216 |
" print('%s Embedding Size: %d' % (col, len(sig)))\n", |
|
|
217 |
"sig_sizes = [len(sig) for sig in sig_names]" |
|
|
218 |
] |
|
|
219 |
}, |
|
|
220 |
{ |
|
|
221 |
"cell_type": "code", |
|
|
222 |
"execution_count": 21, |
|
|
223 |
"metadata": {}, |
|
|
224 |
"outputs": [ |
|
|
225 |
{ |
|
|
226 |
"data": { |
|
|
227 |
"text/plain": [ |
|
|
228 |
"['IFNA10_cnv',\n", |
|
|
229 |
" 'IFNA13_cnv',\n", |
|
|
230 |
" 'IFNA14_cnv',\n", |
|
|
231 |
" 'IFNA16_cnv',\n", |
|
|
232 |
" 'IFNA17_cnv',\n", |
|
|
233 |
" 'IFNA1_cnv',\n", |
|
|
234 |
" 'IFNA21_cnv',\n", |
|
|
235 |
" 'IFNA2_cnv',\n", |
|
|
236 |
" 'IFNA4_cnv',\n", |
|
|
237 |
" 'IFNA5_cnv',\n", |
|
|
238 |
" 'IFNA6_cnv',\n", |
|
|
239 |
" 'IFNA7_cnv',\n", |
|
|
240 |
" 'IFNA8_cnv',\n", |
|
|
241 |
" 'IFNB1_cnv',\n", |
|
|
242 |
" 'IFNE_cnv',\n", |
|
|
243 |
" 'IFNW1_cnv',\n", |
|
|
244 |
" 'PDGFRA_cnv']" |
|
|
245 |
] |
|
|
246 |
}, |
|
|
247 |
"execution_count": 21, |
|
|
248 |
"metadata": {}, |
|
|
249 |
"output_type": "execute_result" |
|
|
250 |
} |
|
|
251 |
], |
|
|
252 |
"source": [ |
|
|
253 |
"sig" |
|
|
254 |
] |
|
|
255 |
}, |
|
|
256 |
{ |
|
|
257 |
"cell_type": "code", |
|
|
258 |
"execution_count": 434, |
|
|
259 |
"metadata": {}, |
|
|
260 |
"outputs": [ |
|
|
261 |
{ |
|
|
262 |
"data": { |
|
|
263 |
"text/html": [ |
|
|
264 |
"<div>\n", |
|
|
265 |
"<style scoped>\n", |
|
|
266 |
" .dataframe tbody tr th:only-of-type {\n", |
|
|
267 |
" vertical-align: middle;\n", |
|
|
268 |
" }\n", |
|
|
269 |
"\n", |
|
|
270 |
" .dataframe tbody tr th {\n", |
|
|
271 |
" vertical-align: top;\n", |
|
|
272 |
" }\n", |
|
|
273 |
"\n", |
|
|
274 |
" .dataframe thead th {\n", |
|
|
275 |
" text-align: right;\n", |
|
|
276 |
" }\n", |
|
|
277 |
"</style>\n", |
|
|
278 |
"<table border=\"1\" class=\"dataframe\">\n", |
|
|
279 |
" <thead>\n", |
|
|
280 |
" <tr style=\"text-align: right;\">\n", |
|
|
281 |
" <th></th>\n", |
|
|
282 |
" <th>NDUFS5_cnv</th>\n", |
|
|
283 |
" <th>MACF1_cnv</th>\n", |
|
|
284 |
" <th>RNA5SP44_cnv</th>\n", |
|
|
285 |
" <th>KIAA0754_cnv</th>\n", |
|
|
286 |
" <th>BMP8A_cnv</th>\n", |
|
|
287 |
" <th>PABPC4_cnv</th>\n", |
|
|
288 |
" <th>SNORA55_cnv</th>\n", |
|
|
289 |
" <th>HEYL_cnv</th>\n", |
|
|
290 |
" <th>HPCAL4_cnv</th>\n", |
|
|
291 |
" <th>NT5C1A_cnv</th>\n", |
|
|
292 |
" <th>...</th>\n", |
|
|
293 |
" <th>ZWINT_rnaseq</th>\n", |
|
|
294 |
" <th>ZXDA_rnaseq</th>\n", |
|
|
295 |
" <th>ZXDB_rnaseq</th>\n", |
|
|
296 |
" <th>ZXDC_rnaseq</th>\n", |
|
|
297 |
" <th>ZYG11A_rnaseq</th>\n", |
|
|
298 |
" <th>ZYG11B_rnaseq</th>\n", |
|
|
299 |
" <th>ZYX_rnaseq</th>\n", |
|
|
300 |
" <th>ZZEF1_rnaseq</th>\n", |
|
|
301 |
" <th>ZZZ3_rnaseq</th>\n", |
|
|
302 |
" <th>TPTEP1_rnaseq</th>\n", |
|
|
303 |
" </tr>\n", |
|
|
304 |
" </thead>\n", |
|
|
305 |
" <tbody>\n", |
|
|
306 |
" <tr>\n", |
|
|
307 |
" <th>0</th>\n", |
|
|
308 |
" <td>-1</td>\n", |
|
|
309 |
" <td>-1</td>\n", |
|
|
310 |
" <td>-1</td>\n", |
|
|
311 |
" <td>-1</td>\n", |
|
|
312 |
" <td>-1</td>\n", |
|
|
313 |
" <td>-1</td>\n", |
|
|
314 |
" <td>-1</td>\n", |
|
|
315 |
" <td>-1</td>\n", |
|
|
316 |
" <td>-1</td>\n", |
|
|
317 |
" <td>-1</td>\n", |
|
|
318 |
" <td>...</td>\n", |
|
|
319 |
" <td>-0.8388</td>\n", |
|
|
320 |
" <td>4.1375</td>\n", |
|
|
321 |
" <td>3.9664</td>\n", |
|
|
322 |
" <td>1.8437</td>\n", |
|
|
323 |
" <td>-0.3959</td>\n", |
|
|
324 |
" <td>-0.2561</td>\n", |
|
|
325 |
" <td>-0.2866</td>\n", |
|
|
326 |
" <td>1.8770</td>\n", |
|
|
327 |
" <td>-0.3179</td>\n", |
|
|
328 |
" <td>-0.3633</td>\n", |
|
|
329 |
" </tr>\n", |
|
|
330 |
" <tr>\n", |
|
|
331 |
" <th>1</th>\n", |
|
|
332 |
" <td>2</td>\n", |
|
|
333 |
" <td>2</td>\n", |
|
|
334 |
" <td>2</td>\n", |
|
|
335 |
" <td>2</td>\n", |
|
|
336 |
" <td>2</td>\n", |
|
|
337 |
" <td>2</td>\n", |
|
|
338 |
" <td>2</td>\n", |
|
|
339 |
" <td>2</td>\n", |
|
|
340 |
" <td>2</td>\n", |
|
|
341 |
" <td>2</td>\n", |
|
|
342 |
" <td>...</td>\n", |
|
|
343 |
" <td>-0.1083</td>\n", |
|
|
344 |
" <td>0.3393</td>\n", |
|
|
345 |
" <td>0.2769</td>\n", |
|
|
346 |
" <td>1.7320</td>\n", |
|
|
347 |
" <td>-0.0975</td>\n", |
|
|
348 |
" <td>2.6955</td>\n", |
|
|
349 |
" <td>-0.6741</td>\n", |
|
|
350 |
" <td>1.0323</td>\n", |
|
|
351 |
" <td>1.2766</td>\n", |
|
|
352 |
" <td>-0.3982</td>\n", |
|
|
353 |
" </tr>\n", |
|
|
354 |
" <tr>\n", |
|
|
355 |
" <th>2</th>\n", |
|
|
356 |
" <td>0</td>\n", |
|
|
357 |
" <td>0</td>\n", |
|
|
358 |
" <td>0</td>\n", |
|
|
359 |
" <td>0</td>\n", |
|
|
360 |
" <td>0</td>\n", |
|
|
361 |
" <td>0</td>\n", |
|
|
362 |
" <td>0</td>\n", |
|
|
363 |
" <td>0</td>\n", |
|
|
364 |
" <td>0</td>\n", |
|
|
365 |
" <td>0</td>\n", |
|
|
366 |
" <td>...</td>\n", |
|
|
367 |
" <td>-0.4155</td>\n", |
|
|
368 |
" <td>1.6846</td>\n", |
|
|
369 |
" <td>0.7711</td>\n", |
|
|
370 |
" <td>-0.3061</td>\n", |
|
|
371 |
" <td>-0.5016</td>\n", |
|
|
372 |
" <td>2.8548</td>\n", |
|
|
373 |
" <td>-0.6171</td>\n", |
|
|
374 |
" <td>-0.8608</td>\n", |
|
|
375 |
" <td>-0.0486</td>\n", |
|
|
376 |
" <td>-0.3962</td>\n", |
|
|
377 |
" </tr>\n", |
|
|
378 |
" <tr>\n", |
|
|
379 |
" <th>3</th>\n", |
|
|
380 |
" <td>0</td>\n", |
|
|
381 |
" <td>0</td>\n", |
|
|
382 |
" <td>0</td>\n", |
|
|
383 |
" <td>0</td>\n", |
|
|
384 |
" <td>0</td>\n", |
|
|
385 |
" <td>0</td>\n", |
|
|
386 |
" <td>0</td>\n", |
|
|
387 |
" <td>0</td>\n", |
|
|
388 |
" <td>0</td>\n", |
|
|
389 |
" <td>0</td>\n", |
|
|
390 |
" <td>...</td>\n", |
|
|
391 |
" <td>-0.8143</td>\n", |
|
|
392 |
" <td>0.8344</td>\n", |
|
|
393 |
" <td>1.5075</td>\n", |
|
|
394 |
" <td>3.6068</td>\n", |
|
|
395 |
" <td>-0.5004</td>\n", |
|
|
396 |
" <td>-0.0747</td>\n", |
|
|
397 |
" <td>-0.2185</td>\n", |
|
|
398 |
" <td>-0.4379</td>\n", |
|
|
399 |
" <td>1.6913</td>\n", |
|
|
400 |
" <td>1.7748</td>\n", |
|
|
401 |
" </tr>\n", |
|
|
402 |
" <tr>\n", |
|
|
403 |
" <th>4</th>\n", |
|
|
404 |
" <td>0</td>\n", |
|
|
405 |
" <td>0</td>\n", |
|
|
406 |
" <td>0</td>\n", |
|
|
407 |
" <td>0</td>\n", |
|
|
408 |
" <td>0</td>\n", |
|
|
409 |
" <td>0</td>\n", |
|
|
410 |
" <td>0</td>\n", |
|
|
411 |
" <td>0</td>\n", |
|
|
412 |
" <td>0</td>\n", |
|
|
413 |
" <td>0</td>\n", |
|
|
414 |
" <td>...</td>\n", |
|
|
415 |
" <td>0.0983</td>\n", |
|
|
416 |
" <td>-0.7908</td>\n", |
|
|
417 |
" <td>-0.0053</td>\n", |
|
|
418 |
" <td>-0.0643</td>\n", |
|
|
419 |
" <td>-0.3706</td>\n", |
|
|
420 |
" <td>0.3870</td>\n", |
|
|
421 |
" <td>-0.5589</td>\n", |
|
|
422 |
" <td>-0.5979</td>\n", |
|
|
423 |
" <td>0.0047</td>\n", |
|
|
424 |
" <td>-0.3548</td>\n", |
|
|
425 |
" </tr>\n", |
|
|
426 |
" <tr>\n", |
|
|
427 |
" <th>...</th>\n", |
|
|
428 |
" <td>...</td>\n", |
|
|
429 |
" <td>...</td>\n", |
|
|
430 |
" <td>...</td>\n", |
|
|
431 |
" <td>...</td>\n", |
|
|
432 |
" <td>...</td>\n", |
|
|
433 |
" <td>...</td>\n", |
|
|
434 |
" <td>...</td>\n", |
|
|
435 |
" <td>...</td>\n", |
|
|
436 |
" <td>...</td>\n", |
|
|
437 |
" <td>...</td>\n", |
|
|
438 |
" <td>...</td>\n", |
|
|
439 |
" <td>...</td>\n", |
|
|
440 |
" <td>...</td>\n", |
|
|
441 |
" <td>...</td>\n", |
|
|
442 |
" <td>...</td>\n", |
|
|
443 |
" <td>...</td>\n", |
|
|
444 |
" <td>...</td>\n", |
|
|
445 |
" <td>...</td>\n", |
|
|
446 |
" <td>...</td>\n", |
|
|
447 |
" <td>...</td>\n", |
|
|
448 |
" <td>...</td>\n", |
|
|
449 |
" </tr>\n", |
|
|
450 |
" <tr>\n", |
|
|
451 |
" <th>368</th>\n", |
|
|
452 |
" <td>2</td>\n", |
|
|
453 |
" <td>2</td>\n", |
|
|
454 |
" <td>2</td>\n", |
|
|
455 |
" <td>2</td>\n", |
|
|
456 |
" <td>2</td>\n", |
|
|
457 |
" <td>2</td>\n", |
|
|
458 |
" <td>2</td>\n", |
|
|
459 |
" <td>2</td>\n", |
|
|
460 |
" <td>2</td>\n", |
|
|
461 |
" <td>2</td>\n", |
|
|
462 |
" <td>...</td>\n", |
|
|
463 |
" <td>-0.0291</td>\n", |
|
|
464 |
" <td>-0.1058</td>\n", |
|
|
465 |
" <td>-0.6721</td>\n", |
|
|
466 |
" <td>0.2802</td>\n", |
|
|
467 |
" <td>1.9504</td>\n", |
|
|
468 |
" <td>-0.8784</td>\n", |
|
|
469 |
" <td>0.9506</td>\n", |
|
|
470 |
" <td>0.0607</td>\n", |
|
|
471 |
" <td>1.1883</td>\n", |
|
|
472 |
" <td>-0.3521</td>\n", |
|
|
473 |
" </tr>\n", |
|
|
474 |
" <tr>\n", |
|
|
475 |
" <th>369</th>\n", |
|
|
476 |
" <td>0</td>\n", |
|
|
477 |
" <td>0</td>\n", |
|
|
478 |
" <td>0</td>\n", |
|
|
479 |
" <td>0</td>\n", |
|
|
480 |
" <td>0</td>\n", |
|
|
481 |
" <td>0</td>\n", |
|
|
482 |
" <td>0</td>\n", |
|
|
483 |
" <td>0</td>\n", |
|
|
484 |
" <td>0</td>\n", |
|
|
485 |
" <td>0</td>\n", |
|
|
486 |
" <td>...</td>\n", |
|
|
487 |
" <td>0.0497</td>\n", |
|
|
488 |
" <td>0.3673</td>\n", |
|
|
489 |
" <td>-0.2208</td>\n", |
|
|
490 |
" <td>0.3034</td>\n", |
|
|
491 |
" <td>3.2580</td>\n", |
|
|
492 |
" <td>-0.2089</td>\n", |
|
|
493 |
" <td>1.6053</td>\n", |
|
|
494 |
" <td>-0.8746</td>\n", |
|
|
495 |
" <td>-0.4491</td>\n", |
|
|
496 |
" <td>-0.3450</td>\n", |
|
|
497 |
" </tr>\n", |
|
|
498 |
" <tr>\n", |
|
|
499 |
" <th>370</th>\n", |
|
|
500 |
" <td>1</td>\n", |
|
|
501 |
" <td>1</td>\n", |
|
|
502 |
" <td>1</td>\n", |
|
|
503 |
" <td>1</td>\n", |
|
|
504 |
" <td>1</td>\n", |
|
|
505 |
" <td>1</td>\n", |
|
|
506 |
" <td>1</td>\n", |
|
|
507 |
" <td>1</td>\n", |
|
|
508 |
" <td>1</td>\n", |
|
|
509 |
" <td>1</td>\n", |
|
|
510 |
" <td>...</td>\n", |
|
|
511 |
" <td>0.3822</td>\n", |
|
|
512 |
" <td>-0.7003</td>\n", |
|
|
513 |
" <td>-0.7661</td>\n", |
|
|
514 |
" <td>-1.7035</td>\n", |
|
|
515 |
" <td>-0.5423</td>\n", |
|
|
516 |
" <td>-0.3488</td>\n", |
|
|
517 |
" <td>1.3713</td>\n", |
|
|
518 |
" <td>-0.4365</td>\n", |
|
|
519 |
" <td>2.3456</td>\n", |
|
|
520 |
" <td>-0.3866</td>\n", |
|
|
521 |
" </tr>\n", |
|
|
522 |
" <tr>\n", |
|
|
523 |
" <th>371</th>\n", |
|
|
524 |
" <td>0</td>\n", |
|
|
525 |
" <td>0</td>\n", |
|
|
526 |
" <td>0</td>\n", |
|
|
527 |
" <td>0</td>\n", |
|
|
528 |
" <td>0</td>\n", |
|
|
529 |
" <td>0</td>\n", |
|
|
530 |
" <td>0</td>\n", |
|
|
531 |
" <td>0</td>\n", |
|
|
532 |
" <td>0</td>\n", |
|
|
533 |
" <td>0</td>\n", |
|
|
534 |
" <td>...</td>\n", |
|
|
535 |
" <td>-0.6853</td>\n", |
|
|
536 |
" <td>-1.0240</td>\n", |
|
|
537 |
" <td>-1.2890</td>\n", |
|
|
538 |
" <td>-1.5666</td>\n", |
|
|
539 |
" <td>-0.1270</td>\n", |
|
|
540 |
" <td>-1.4662</td>\n", |
|
|
541 |
" <td>0.3981</td>\n", |
|
|
542 |
" <td>-0.5976</td>\n", |
|
|
543 |
" <td>-1.3822</td>\n", |
|
|
544 |
" <td>-0.4157</td>\n", |
|
|
545 |
" </tr>\n", |
|
|
546 |
" <tr>\n", |
|
|
547 |
" <th>372</th>\n", |
|
|
548 |
" <td>0</td>\n", |
|
|
549 |
" <td>0</td>\n", |
|
|
550 |
" <td>0</td>\n", |
|
|
551 |
" <td>0</td>\n", |
|
|
552 |
" <td>0</td>\n", |
|
|
553 |
" <td>0</td>\n", |
|
|
554 |
" <td>0</td>\n", |
|
|
555 |
" <td>0</td>\n", |
|
|
556 |
" <td>0</td>\n", |
|
|
557 |
" <td>0</td>\n", |
|
|
558 |
" <td>...</td>\n", |
|
|
559 |
" <td>0.0517</td>\n", |
|
|
560 |
" <td>-0.3570</td>\n", |
|
|
561 |
" <td>-0.4843</td>\n", |
|
|
562 |
" <td>-0.3792</td>\n", |
|
|
563 |
" <td>-0.1964</td>\n", |
|
|
564 |
" <td>0.4200</td>\n", |
|
|
565 |
" <td>3.2547</td>\n", |
|
|
566 |
" <td>-0.1232</td>\n", |
|
|
567 |
" <td>3.4519</td>\n", |
|
|
568 |
" <td>-0.1962</td>\n", |
|
|
569 |
" </tr>\n", |
|
|
570 |
" </tbody>\n", |
|
|
571 |
"</table>\n", |
|
|
572 |
"<p>373 rows × 20395 columns</p>\n", |
|
|
573 |
"</div>" |
|
|
574 |
], |
|
|
575 |
"text/plain": [ |
|
|
576 |
" NDUFS5_cnv MACF1_cnv RNA5SP44_cnv KIAA0754_cnv BMP8A_cnv PABPC4_cnv \\\n", |
|
|
577 |
"0 -1 -1 -1 -1 -1 -1 \n", |
|
|
578 |
"1 2 2 2 2 2 2 \n", |
|
|
579 |
"2 0 0 0 0 0 0 \n", |
|
|
580 |
"3 0 0 0 0 0 0 \n", |
|
|
581 |
"4 0 0 0 0 0 0 \n", |
|
|
582 |
".. ... ... ... ... ... ... \n", |
|
|
583 |
"368 2 2 2 2 2 2 \n", |
|
|
584 |
"369 0 0 0 0 0 0 \n", |
|
|
585 |
"370 1 1 1 1 1 1 \n", |
|
|
586 |
"371 0 0 0 0 0 0 \n", |
|
|
587 |
"372 0 0 0 0 0 0 \n", |
|
|
588 |
"\n", |
|
|
589 |
" SNORA55_cnv HEYL_cnv HPCAL4_cnv NT5C1A_cnv ... ZWINT_rnaseq \\\n", |
|
|
590 |
"0 -1 -1 -1 -1 ... -0.8388 \n", |
|
|
591 |
"1 2 2 2 2 ... -0.1083 \n", |
|
|
592 |
"2 0 0 0 0 ... -0.4155 \n", |
|
|
593 |
"3 0 0 0 0 ... -0.8143 \n", |
|
|
594 |
"4 0 0 0 0 ... 0.0983 \n", |
|
|
595 |
".. ... ... ... ... ... ... \n", |
|
|
596 |
"368 2 2 2 2 ... -0.0291 \n", |
|
|
597 |
"369 0 0 0 0 ... 0.0497 \n", |
|
|
598 |
"370 1 1 1 1 ... 0.3822 \n", |
|
|
599 |
"371 0 0 0 0 ... -0.6853 \n", |
|
|
600 |
"372 0 0 0 0 ... 0.0517 \n", |
|
|
601 |
"\n", |
|
|
602 |
" ZXDA_rnaseq ZXDB_rnaseq ZXDC_rnaseq ZYG11A_rnaseq ZYG11B_rnaseq \\\n", |
|
|
603 |
"0 4.1375 3.9664 1.8437 -0.3959 -0.2561 \n", |
|
|
604 |
"1 0.3393 0.2769 1.7320 -0.0975 2.6955 \n", |
|
|
605 |
"2 1.6846 0.7711 -0.3061 -0.5016 2.8548 \n", |
|
|
606 |
"3 0.8344 1.5075 3.6068 -0.5004 -0.0747 \n", |
|
|
607 |
"4 -0.7908 -0.0053 -0.0643 -0.3706 0.3870 \n", |
|
|
608 |
".. ... ... ... ... ... \n", |
|
|
609 |
"368 -0.1058 -0.6721 0.2802 1.9504 -0.8784 \n", |
|
|
610 |
"369 0.3673 -0.2208 0.3034 3.2580 -0.2089 \n", |
|
|
611 |
"370 -0.7003 -0.7661 -1.7035 -0.5423 -0.3488 \n", |
|
|
612 |
"371 -1.0240 -1.2890 -1.5666 -0.1270 -1.4662 \n", |
|
|
613 |
"372 -0.3570 -0.4843 -0.3792 -0.1964 0.4200 \n", |
|
|
614 |
"\n", |
|
|
615 |
" ZYX_rnaseq ZZEF1_rnaseq ZZZ3_rnaseq TPTEP1_rnaseq \n", |
|
|
616 |
"0 -0.2866 1.8770 -0.3179 -0.3633 \n", |
|
|
617 |
"1 -0.6741 1.0323 1.2766 -0.3982 \n", |
|
|
618 |
"2 -0.6171 -0.8608 -0.0486 -0.3962 \n", |
|
|
619 |
"3 -0.2185 -0.4379 1.6913 1.7748 \n", |
|
|
620 |
"4 -0.5589 -0.5979 0.0047 -0.3548 \n", |
|
|
621 |
".. ... ... ... ... \n", |
|
|
622 |
"368 0.9506 0.0607 1.1883 -0.3521 \n", |
|
|
623 |
"369 1.6053 -0.8746 -0.4491 -0.3450 \n", |
|
|
624 |
"370 1.3713 -0.4365 2.3456 -0.3866 \n", |
|
|
625 |
"371 0.3981 -0.5976 -1.3822 -0.4157 \n", |
|
|
626 |
"372 3.2547 -0.1232 3.4519 -0.1962 \n", |
|
|
627 |
"\n", |
|
|
628 |
"[373 rows x 20395 columns]" |
|
|
629 |
] |
|
|
630 |
}, |
|
|
631 |
"execution_count": 434, |
|
|
632 |
"metadata": {}, |
|
|
633 |
"output_type": "execute_result" |
|
|
634 |
} |
|
|
635 |
], |
|
|
636 |
"source": [ |
|
|
637 |
"genomic_features" |
|
|
638 |
] |
|
|
639 |
}, |
|
|
640 |
{ |
|
|
641 |
"cell_type": "code", |
|
|
642 |
"execution_count": 2, |
|
|
643 |
"metadata": {}, |
|
|
644 |
"outputs": [], |
|
|
645 |
"source": [ |
|
|
646 |
"import torch\n", |
|
|
647 |
"import torch.nn as nn\n", |
|
|
648 |
"import torch.nn.functional as F\n", |
|
|
649 |
"import pdb\n", |
|
|
650 |
"import numpy as np\n", |
|
|
651 |
"\n", |
|
|
652 |
"class MIL_Sum_FC_surv(nn.Module):\n", |
|
|
653 |
" def __init__(self, size_arg = \"small\", dropout=0.25, n_classes=4):\n", |
|
|
654 |
" super(MIL_Sum_FC_surv, self).__init__()\n", |
|
|
655 |
"\n", |
|
|
656 |
" self.size_dict = {\"small\": [1024, 512, 256], \"big\": [1024, 512, 384]}\n", |
|
|
657 |
" size = self.size_dict[size_arg]\n", |
|
|
658 |
" self.phi = nn.Sequential(*[nn.Linear(size[0], size[1]), nn.ReLU(), nn.Dropout(dropout)])\n", |
|
|
659 |
" self.rho = nn.Sequential(*[nn.Linear(size[1], size[2]), nn.ReLU(), nn.Dropout(dropout)])\n", |
|
|
660 |
" self.classifier = nn.Linear(size[2], n_classes)\n", |
|
|
661 |
"\n", |
|
|
662 |
" def relocate(self):\n", |
|
|
663 |
" device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n", |
|
|
664 |
" if torch.cuda.device_count() >= 1:\n", |
|
|
665 |
" device_ids = list(range(torch.cuda.device_count()))\n", |
|
|
666 |
" self.phi = nn.DataParallel(self.phi, device_ids=device_ids).to('cuda:0')\n", |
|
|
667 |
"\n", |
|
|
668 |
" self.rho = self.rho.to(device)\n", |
|
|
669 |
" self.classifier = self.classifier.to(device)\n", |
|
|
670 |
"\n", |
|
|
671 |
" def forward(self, **kwargs):\n", |
|
|
672 |
" h = kwargs['x_path']\n", |
|
|
673 |
"\n", |
|
|
674 |
" h = self.phi(h).sum(axis=0)\n", |
|
|
675 |
" h = self.rho(h)\n", |
|
|
676 |
" logits = self.classifier(h).unsqueeze(0)\n", |
|
|
677 |
" Y_hat = torch.topk(logits, 1, dim = 1)[1]\n", |
|
|
678 |
" hazards = torch.sigmoid(logits)\n", |
|
|
679 |
" S = torch.cumprod(1 - hazards, dim=1)\n", |
|
|
680 |
" \n", |
|
|
681 |
" return hazards, S, Y_hat, None, None\n", |
|
|
682 |
"\n", |
|
|
683 |
"from os.path import join\n", |
|
|
684 |
"from collections import OrderedDict\n", |
|
|
685 |
"\n", |
|
|
686 |
"import torch\n", |
|
|
687 |
"import torch.nn as nn\n", |
|
|
688 |
"import torch.nn.functional as F\n", |
|
|
689 |
"import pdb\n", |
|
|
690 |
"import numpy as np\n", |
|
|
691 |
"\n", |
|
|
692 |
"\"\"\"\n", |
|
|
693 |
"A Modified Implementation of Deep Attention MIL\n", |
|
|
694 |
"\"\"\"\n", |
|
|
695 |
"\n", |
|
|
696 |
"\n", |
|
|
697 |
"\"\"\"\n", |
|
|
698 |
"Attention Network without Gating (2 fc layers)\n", |
|
|
699 |
"args:\n", |
|
|
700 |
" L: input feature dimension\n", |
|
|
701 |
" D: hidden layer dimension\n", |
|
|
702 |
" dropout: whether to use dropout (p = 0.25)\n", |
|
|
703 |
" n_classes: number of classes (experimental usage for multiclass MIL)\n", |
|
|
704 |
"\"\"\"\n", |
|
|
705 |
"class Attn_Net(nn.Module):\n", |
|
|
706 |
"\n", |
|
|
707 |
" def __init__(self, L = 1024, D = 256, dropout = False, n_classes = 1):\n", |
|
|
708 |
" super(Attn_Net, self).__init__()\n", |
|
|
709 |
" self.module = [\n", |
|
|
710 |
" nn.Linear(L, D),\n", |
|
|
711 |
" nn.Tanh()]\n", |
|
|
712 |
"\n", |
|
|
713 |
" if dropout:\n", |
|
|
714 |
" self.module.append(nn.Dropout(0.25))\n", |
|
|
715 |
"\n", |
|
|
716 |
" self.module.append(nn.Linear(D, n_classes))\n", |
|
|
717 |
" \n", |
|
|
718 |
" self.module = nn.Sequential(*self.module)\n", |
|
|
719 |
" \n", |
|
|
720 |
" def forward(self, x):\n", |
|
|
721 |
" return self.module(x), x # N x n_classes\n", |
|
|
722 |
"\n", |
|
|
723 |
"\"\"\"\n", |
|
|
724 |
"Attention Network with Sigmoid Gating (3 fc layers)\n", |
|
|
725 |
"args:\n", |
|
|
726 |
" L: input feature dimension\n", |
|
|
727 |
" D: hidden layer dimension\n", |
|
|
728 |
" dropout: whether to use dropout (p = 0.25)\n", |
|
|
729 |
" n_classes: number of classes (experimental usage for multiclass MIL)\n", |
|
|
730 |
"\"\"\"\n", |
|
|
731 |
"class Attn_Net_Gated(nn.Module):\n", |
|
|
732 |
"\n", |
|
|
733 |
" def __init__(self, L = 1024, D = 256, dropout = False, n_classes = 1):\n", |
|
|
734 |
" super(Attn_Net_Gated, self).__init__()\n", |
|
|
735 |
" self.attention_a = [\n", |
|
|
736 |
" nn.Linear(L, D),\n", |
|
|
737 |
" nn.Tanh()]\n", |
|
|
738 |
" \n", |
|
|
739 |
" self.attention_b = [nn.Linear(L, D),\n", |
|
|
740 |
" nn.Sigmoid()]\n", |
|
|
741 |
" if dropout:\n", |
|
|
742 |
" self.attention_a.append(nn.Dropout(0.25))\n", |
|
|
743 |
" self.attention_b.append(nn.Dropout(0.25))\n", |
|
|
744 |
"\n", |
|
|
745 |
" self.attention_a = nn.Sequential(*self.attention_a)\n", |
|
|
746 |
" self.attention_b = nn.Sequential(*self.attention_b)\n", |
|
|
747 |
" \n", |
|
|
748 |
" self.attention_c = nn.Linear(D, n_classes)\n", |
|
|
749 |
"\n", |
|
|
750 |
" def forward(self, x):\n", |
|
|
751 |
" a = self.attention_a(x)\n", |
|
|
752 |
" b = self.attention_b(x)\n", |
|
|
753 |
" A = a.mul(b)\n", |
|
|
754 |
" A = self.attention_c(A) # N x n_classes\n", |
|
|
755 |
" return A, x\n", |
|
|
756 |
" \n", |
|
|
757 |
"class MIL_Cluster_FC_surv(nn.Module):\n", |
|
|
758 |
" def __init__(self, num_clusters=10, size_arg = \"small\", dropout=0.25, n_classes=4):\n", |
|
|
759 |
" super(MIL_Cluster_FC_surv, self).__init__()\n", |
|
|
760 |
" self.size_dict = {\"small\": [1024, 512, 256], \"big\": [1024, 512, 384]}\n", |
|
|
761 |
" self.num_clusters = num_clusters\n", |
|
|
762 |
" \n", |
|
|
763 |
" ### Phenotype Learning\n", |
|
|
764 |
" size = self.size_dict[size_arg]\n", |
|
|
765 |
" phis = []\n", |
|
|
766 |
" for phenotype_i in range(num_clusters):\n", |
|
|
767 |
" phi = [nn.Linear(size[0], size[1]), nn.ReLU(), nn.Dropout(dropout),\n", |
|
|
768 |
" nn.Linear(size[1], size[1]), nn.ReLU(), nn.Dropout(dropout)]\n", |
|
|
769 |
" phis.append(nn.Sequential(*phi))\n", |
|
|
770 |
" self.phis = nn.ModuleList(phis)\n", |
|
|
771 |
" self.pool1d = nn.AdaptiveAvgPool1d(1)\n", |
|
|
772 |
" \n", |
|
|
773 |
" \n", |
|
|
774 |
" ### WSI Attention MIL Construction\n", |
|
|
775 |
" fc = [nn.Linear(size[1], size[1]), nn.ReLU()]\n", |
|
|
776 |
" fc.append(nn.Dropout(0.25))\n", |
|
|
777 |
" attention_net = Attn_Net_Gated(L=size[1], D=size[2], dropout=dropout, n_classes=1)\n", |
|
|
778 |
" fc.append(attention_net)\n", |
|
|
779 |
" self.attention_net = nn.Sequential(*fc)\n", |
|
|
780 |
"\n", |
|
|
781 |
" \n", |
|
|
782 |
" self.rho = nn.Sequential(*[nn.Linear(size[1], size[2]), nn.ReLU(), nn.Dropout(dropout)])\n", |
|
|
783 |
" self.classifier = nn.Linear(size[2], n_classes)\n", |
|
|
784 |
"\n", |
|
|
785 |
" def relocate(self):\n", |
|
|
786 |
" device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n", |
|
|
787 |
" if torch.cuda.device_count() >= 1:\n", |
|
|
788 |
" device_ids = list(range(torch.cuda.device_count()))\n", |
|
|
789 |
" self.phis = nn.DataParallel(self.phi, device_ids=device_ids).to('cuda:0')\n", |
|
|
790 |
"\n", |
|
|
791 |
" self.rho = self.rho.to(device)\n", |
|
|
792 |
" self.classifier = self.classifier.to(device)\n", |
|
|
793 |
"\n", |
|
|
794 |
" def forward(self, **kwargs):\n", |
|
|
795 |
" x_path = kwargs['x_path']\n", |
|
|
796 |
" ### Phenotyping\n", |
|
|
797 |
" h_phenotypes = []\n", |
|
|
798 |
" from sklearn.cluster import KMeans\n", |
|
|
799 |
" kmeans = KMeans(n_clusters=self.num_clusters, random_state=2021).fit(X)\n", |
|
|
800 |
" #cluster_ids_x, cluster_centers = kmeans(X=x_path, num_clusters=self.num_clusters, distance='euclidean', device=torch.device('cpu'))\n", |
|
|
801 |
" cluster_ids_x = KMeans(n_clusters=10, random_state=2021, max_iter=20).fit_predict(x_path)\n", |
|
|
802 |
" for i in range(self.num_clusters):\n", |
|
|
803 |
" h_phenotypes_i = self.phis[i](x_path[cluster_ids_x==i])\n", |
|
|
804 |
" h_phenotypes.append(self.pool1d(h_phenotypes_i.T.unsqueeze(0)).squeeze(2))\n", |
|
|
805 |
" h_phenotypes = torch.stack(h_phenotypes, dim=1).squeeze(0)\n", |
|
|
806 |
"\n", |
|
|
807 |
"\n", |
|
|
808 |
" ### Attention MIL\n", |
|
|
809 |
" A, h = self.attention_net(h_phenotypes) \n", |
|
|
810 |
" A = torch.transpose(A, 1, 0)\n", |
|
|
811 |
" if 'attention_only' in kwargs.keys():\n", |
|
|
812 |
" if kwargs['attention_only']:\n", |
|
|
813 |
" return A\n", |
|
|
814 |
" A_raw = A \n", |
|
|
815 |
" A = F.softmax(A, dim=1) \n", |
|
|
816 |
" h = torch.mm(A, h_phenotypes)\n", |
|
|
817 |
"\n", |
|
|
818 |
" \n", |
|
|
819 |
" h = self.rho(h)\n", |
|
|
820 |
" logits = self.classifier(h).unsqueeze(0)\n", |
|
|
821 |
" Y_hat = torch.topk(logits, 1, dim = 1)[1]\n", |
|
|
822 |
" hazards = torch.sigmoid(logits)\n", |
|
|
823 |
" S = torch.cumprod(1 - hazards, dim=1)\n", |
|
|
824 |
" \n", |
|
|
825 |
" return hazards, S, Y_hat, None, None" |
|
|
826 |
] |
|
|
827 |
}, |
|
|
828 |
{ |
|
|
829 |
"cell_type": "code", |
|
|
830 |
"execution_count": 15, |
|
|
831 |
"metadata": {}, |
|
|
832 |
"outputs": [], |
|
|
833 |
"source": [ |
|
|
834 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.cuda.FloatTensor)\n" |
|
|
835 |
] |
|
|
836 |
}, |
|
|
837 |
{ |
|
|
838 |
"cell_type": "code", |
|
|
839 |
"execution_count": 17, |
|
|
840 |
"metadata": {}, |
|
|
841 |
"outputs": [], |
|
|
842 |
"source": [ |
|
|
843 |
"from sklearn.cluster import KMeans\n", |
|
|
844 |
"kmeans = KMeans(n_clusters=10, random_state=2021, max_iter=20).fit_predict(x_path.cpu())" |
|
|
845 |
] |
|
|
846 |
}, |
|
|
847 |
{ |
|
|
848 |
"cell_type": "code", |
|
|
849 |
"execution_count": 18, |
|
|
850 |
"metadata": {}, |
|
|
851 |
"outputs": [ |
|
|
852 |
{ |
|
|
853 |
"data": { |
|
|
854 |
"text/plain": [ |
|
|
855 |
"array([5, 5, 3, 5, 8, 4, 8, 7, 5, 4, 9, 1, 9, 1, 6, 1, 1, 0, 5, 0, 4, 3,\n", |
|
|
856 |
" 0, 6, 3, 1, 0, 7, 9, 8, 0, 5, 5, 3, 0, 1, 5, 1, 0, 6, 6, 4, 1, 5,\n", |
|
|
857 |
" 3, 0, 1, 0, 8, 5, 1, 8, 1, 0, 5, 0, 2, 5, 6, 5, 0, 0, 5, 1, 2, 7,\n", |
|
|
858 |
" 4, 6, 5, 3, 0, 7, 9, 1, 3, 4, 4, 5, 7, 9, 9, 5, 0, 1, 9, 1, 2, 0,\n", |
|
|
859 |
" 6, 3, 1, 1, 2, 4, 0, 5, 1, 1, 1, 0, 0, 9, 8, 1, 5, 5, 0, 9, 2, 3,\n", |
|
|
860 |
" 7, 0, 1, 6, 7, 5, 3, 5, 0, 1, 6, 1, 6, 2, 8, 7, 6, 1, 6, 2, 5, 0,\n", |
|
|
861 |
" 1, 6, 0, 9, 2, 1, 0, 1, 7, 7, 6, 1, 6, 0, 3, 4, 1, 3, 2, 4, 4, 5,\n", |
|
|
862 |
" 4, 1, 1, 9, 6, 0, 3, 6, 4, 8, 7, 9, 6, 5, 5, 9, 0, 6, 0, 1, 9, 2,\n", |
|
|
863 |
" 3, 5, 1, 9, 6, 1, 0, 6, 6, 0, 0, 6, 7, 1, 6, 1, 1, 1, 4, 0, 2, 1,\n", |
|
|
864 |
" 9, 5, 7, 5, 9, 0, 1, 0, 6, 2, 2, 1, 1, 5, 3, 5, 3, 6, 5, 6, 9, 5,\n", |
|
|
865 |
" 2, 2, 2, 6, 0, 0, 0, 5, 2, 6, 6, 0, 2, 5, 1, 9, 2, 4, 4, 0, 4, 7,\n", |
|
|
866 |
" 4, 1, 1, 3, 6, 0, 1, 2, 4, 0, 8, 1, 8, 5, 5, 7, 4, 1, 6, 1, 0, 8,\n", |
|
|
867 |
" 6, 1, 1, 4, 8, 7, 5, 2, 3, 0, 2, 9, 5, 6, 4, 3, 6, 5, 5, 4, 6, 6,\n", |
|
|
868 |
" 0, 1, 5, 1, 1, 1, 1, 9, 5, 7, 3, 0, 2, 4, 0, 5, 4, 0, 5, 0, 6, 0,\n", |
|
|
869 |
" 3, 1, 4, 6, 3, 7, 1, 6, 7, 0, 1, 4, 6, 1, 6, 0, 6, 0, 5, 9, 1, 1,\n", |
|
|
870 |
" 3, 1, 5, 6, 1, 6, 6, 8, 2, 0, 7, 9, 9, 6, 0, 6, 2, 6, 8, 0, 8, 5,\n", |
|
|
871 |
" 1, 3, 1, 9, 2, 3, 5, 8, 2, 5, 6, 6, 5, 2, 9, 0, 1, 8, 5, 9, 5, 1,\n", |
|
|
872 |
" 0, 1, 0, 8, 6, 1, 7, 2, 8, 3, 1, 6, 2, 2, 1, 6, 0, 2, 6, 1, 1, 4,\n", |
|
|
873 |
" 5, 6, 4, 0, 5, 0, 9, 0, 4, 8, 0, 7, 6, 5, 5, 0, 4, 1, 1, 2, 2, 0,\n", |
|
|
874 |
" 0, 6, 4, 0, 7, 7, 2, 3, 1, 4, 7, 9, 4, 7, 2, 4, 5, 6, 4, 5, 7, 9,\n", |
|
|
875 |
" 8, 0, 6, 2, 0, 6, 6, 3, 5, 4, 4, 0, 1, 0, 5, 3, 1, 6, 0, 7, 4, 1,\n", |
|
|
876 |
" 6, 3, 6, 0, 4, 1, 5, 7, 3, 1, 4, 8, 0, 7, 0, 6, 1, 1, 0, 1, 5, 1,\n", |
|
|
877 |
" 2, 3, 2, 3, 8, 8, 4, 6, 5, 6, 1, 0, 7, 6, 4, 4], dtype=int32)" |
|
|
878 |
] |
|
|
879 |
}, |
|
|
880 |
"execution_count": 18, |
|
|
881 |
"metadata": {}, |
|
|
882 |
"output_type": "execute_result" |
|
|
883 |
} |
|
|
884 |
], |
|
|
885 |
"source": [ |
|
|
886 |
"kmeans" |
|
|
887 |
] |
|
|
888 |
}, |
|
|
889 |
{ |
|
|
890 |
"cell_type": "code", |
|
|
891 |
"execution_count": 2, |
|
|
892 |
"metadata": {}, |
|
|
893 |
"outputs": [ |
|
|
894 |
{ |
|
|
895 |
"data": { |
|
|
896 |
"text/plain": [ |
|
|
897 |
"(tensor([[0.9992, 0.0000, 0.0000, 1.0000]], grad_fn=<SigmoidBackward>),\n", |
|
|
898 |
" tensor([[0.0008, 0.0008, 0.0008, 0.0000]], grad_fn=<CumprodBackward>),\n", |
|
|
899 |
" tensor([[3]]),\n", |
|
|
900 |
" None,\n", |
|
|
901 |
" None)" |
|
|
902 |
] |
|
|
903 |
}, |
|
|
904 |
"execution_count": 2, |
|
|
905 |
"metadata": {}, |
|
|
906 |
"output_type": "execute_result" |
|
|
907 |
} |
|
|
908 |
], |
|
|
909 |
"source": [ |
|
|
910 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.FloatTensor)\n", |
|
|
911 |
"model = MIL_Sum_FC_surv()\n", |
|
|
912 |
"model.forward(x_path=x_path)" |
|
|
913 |
] |
|
|
914 |
}, |
|
|
915 |
{ |
|
|
916 |
"cell_type": "code", |
|
|
917 |
"execution_count": 3, |
|
|
918 |
"metadata": {}, |
|
|
919 |
"outputs": [ |
|
|
920 |
{ |
|
|
921 |
"data": { |
|
|
922 |
"text/plain": [ |
|
|
923 |
"(tensor([[4.2595e-07, 1.0000e+00, 0.0000e+00, 7.2488e-12]],\n", |
|
|
924 |
" grad_fn=<SigmoidBackward>),\n", |
|
|
925 |
" tensor([[1.0000, 0.0000, 0.0000, 0.0000]], grad_fn=<CumprodBackward>),\n", |
|
|
926 |
" tensor([[1]]),\n", |
|
|
927 |
" None,\n", |
|
|
928 |
" None)" |
|
|
929 |
] |
|
|
930 |
}, |
|
|
931 |
"execution_count": 3, |
|
|
932 |
"metadata": {}, |
|
|
933 |
"output_type": "execute_result" |
|
|
934 |
} |
|
|
935 |
], |
|
|
936 |
"source": [ |
|
|
937 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.FloatTensor)\n", |
|
|
938 |
"self = MIL_Cluster_FC_surv()\n", |
|
|
939 |
"model.forward(x_path=x_path)" |
|
|
940 |
] |
|
|
941 |
}, |
|
|
942 |
{ |
|
|
943 |
"cell_type": "code", |
|
|
944 |
"execution_count": 7, |
|
|
945 |
"metadata": {}, |
|
|
946 |
"outputs": [], |
|
|
947 |
"source": [ |
|
|
948 |
"import os\n", |
|
|
949 |
"fname = os.path.join('/media/ssd1/pan-cancer/tcga_gbm_20x_features/h5_files/TCGA-02-0001-01Z-00-DX1.83fce43e-42ac-4dcd-b156-2908e75f2e47.h5')" |
|
|
950 |
] |
|
|
951 |
}, |
|
|
952 |
{ |
|
|
953 |
"cell_type": "code", |
|
|
954 |
"execution_count": 27, |
|
|
955 |
"metadata": {}, |
|
|
956 |
"outputs": [], |
|
|
957 |
"source": [ |
|
|
958 |
"import h5py\n", |
|
|
959 |
"h5 = h5py.File(fname, \"r\")\n", |
|
|
960 |
"coords = np.array(h5['coords'])" |
|
|
961 |
] |
|
|
962 |
}, |
|
|
963 |
{ |
|
|
964 |
"cell_type": "code", |
|
|
965 |
"execution_count": null, |
|
|
966 |
"metadata": {}, |
|
|
967 |
"outputs": [], |
|
|
968 |
"source": [ |
|
|
969 |
"fm" |
|
|
970 |
] |
|
|
971 |
}, |
|
|
972 |
{ |
|
|
973 |
"cell_type": "code", |
|
|
974 |
"execution_count": 17, |
|
|
975 |
"metadata": {}, |
|
|
976 |
"outputs": [ |
|
|
977 |
{ |
|
|
978 |
"data": { |
|
|
979 |
"text/plain": [ |
|
|
980 |
"array([43121, 29428])" |
|
|
981 |
] |
|
|
982 |
}, |
|
|
983 |
"execution_count": 17, |
|
|
984 |
"metadata": {}, |
|
|
985 |
"output_type": "execute_result" |
|
|
986 |
} |
|
|
987 |
], |
|
|
988 |
"source": [ |
|
|
989 |
"np.array(h5['coords'])[0]" |
|
|
990 |
] |
|
|
991 |
}, |
|
|
992 |
{ |
|
|
993 |
"cell_type": "code", |
|
|
994 |
"execution_count": 19, |
|
|
995 |
"metadata": {}, |
|
|
996 |
"outputs": [ |
|
|
997 |
{ |
|
|
998 |
"data": { |
|
|
999 |
"text/plain": [ |
|
|
1000 |
"array([43121, 29940])" |
|
|
1001 |
] |
|
|
1002 |
}, |
|
|
1003 |
"execution_count": 19, |
|
|
1004 |
"metadata": {}, |
|
|
1005 |
"output_type": "execute_result" |
|
|
1006 |
} |
|
|
1007 |
], |
|
|
1008 |
"source": [ |
|
|
1009 |
"np.array(h5['coords'])[1]" |
|
|
1010 |
] |
|
|
1011 |
}, |
|
|
1012 |
{ |
|
|
1013 |
"cell_type": "code", |
|
|
1014 |
"execution_count": 20, |
|
|
1015 |
"metadata": {}, |
|
|
1016 |
"outputs": [ |
|
|
1017 |
{ |
|
|
1018 |
"data": { |
|
|
1019 |
"text/plain": [ |
|
|
1020 |
"512" |
|
|
1021 |
] |
|
|
1022 |
}, |
|
|
1023 |
"execution_count": 20, |
|
|
1024 |
"metadata": {}, |
|
|
1025 |
"output_type": "execute_result" |
|
|
1026 |
} |
|
|
1027 |
], |
|
|
1028 |
"source": [ |
|
|
1029 |
"np.array(h5['coords'])[1][1] - np.array(h5['coords'])[0][1]" |
|
|
1030 |
] |
|
|
1031 |
}, |
|
|
1032 |
{ |
|
|
1033 |
"cell_type": "code", |
|
|
1034 |
"execution_count": 21, |
|
|
1035 |
"metadata": {}, |
|
|
1036 |
"outputs": [ |
|
|
1037 |
{ |
|
|
1038 |
"data": { |
|
|
1039 |
"text/plain": [ |
|
|
1040 |
"512" |
|
|
1041 |
] |
|
|
1042 |
}, |
|
|
1043 |
"execution_count": 21, |
|
|
1044 |
"metadata": {}, |
|
|
1045 |
"output_type": "execute_result" |
|
|
1046 |
} |
|
|
1047 |
], |
|
|
1048 |
"source": [ |
|
|
1049 |
"np.array(h5['coords'])[2][1] - np.array(h5['coords'])[1][1]" |
|
|
1050 |
] |
|
|
1051 |
}, |
|
|
1052 |
{ |
|
|
1053 |
"cell_type": "code", |
|
|
1054 |
"execution_count": 23, |
|
|
1055 |
"metadata": {}, |
|
|
1056 |
"outputs": [], |
|
|
1057 |
"source": [ |
|
|
1058 |
"import nmslib\n", |
|
|
1059 |
"class Hnsw:\n", |
|
|
1060 |
"\n", |
|
|
1061 |
" def __init__(self, space='cosinesimil', index_params=None,\n", |
|
|
1062 |
" query_params=None, print_progress=True):\n", |
|
|
1063 |
" self.space = space\n", |
|
|
1064 |
" self.index_params = index_params\n", |
|
|
1065 |
" self.query_params = query_params\n", |
|
|
1066 |
" self.print_progress = print_progress\n", |
|
|
1067 |
"\n", |
|
|
1068 |
" def fit(self, X):\n", |
|
|
1069 |
" index_params = self.index_params\n", |
|
|
1070 |
" if index_params is None:\n", |
|
|
1071 |
" index_params = {'M': 16, 'post': 0, 'efConstruction': 400}\n", |
|
|
1072 |
"\n", |
|
|
1073 |
" query_params = self.query_params\n", |
|
|
1074 |
" if query_params is None:\n", |
|
|
1075 |
" query_params = {'ef': 90}\n", |
|
|
1076 |
"\n", |
|
|
1077 |
" # this is the actual nmslib part, hopefully the syntax should\n", |
|
|
1078 |
" # be pretty readable, the documentation also has a more verbiage\n", |
|
|
1079 |
" # introduction: https://nmslib.github.io/nmslib/quickstart.html\n", |
|
|
1080 |
" index = nmslib.init(space=self.space, method='hnsw')\n", |
|
|
1081 |
" index.addDataPointBatch(X)\n", |
|
|
1082 |
" index.createIndex(index_params, print_progress=self.print_progress)\n", |
|
|
1083 |
" index.setQueryTimeParams(query_params)\n", |
|
|
1084 |
"\n", |
|
|
1085 |
" self.index_ = index\n", |
|
|
1086 |
" self.index_params_ = index_params\n", |
|
|
1087 |
" self.query_params_ = query_params\n", |
|
|
1088 |
" return self\n", |
|
|
1089 |
"\n", |
|
|
1090 |
" def query(self, vector, topn):\n", |
|
|
1091 |
" # the knnQuery returns indices and corresponding distance\n", |
|
|
1092 |
" # we will throw the distance away for now\n", |
|
|
1093 |
" indices, _ = self.index_.knnQuery(vector, k=topn)\n", |
|
|
1094 |
" return indices" |
|
|
1095 |
] |
|
|
1096 |
}, |
|
|
1097 |
{ |
|
|
1098 |
"cell_type": "code", |
|
|
1099 |
"execution_count": null, |
|
|
1100 |
"metadata": {}, |
|
|
1101 |
"outputs": [], |
|
|
1102 |
"source": [ |
|
|
1103 |
"x" |
|
|
1104 |
] |
|
|
1105 |
}, |
|
|
1106 |
{ |
|
|
1107 |
"cell_type": "code", |
|
|
1108 |
"execution_count": 54, |
|
|
1109 |
"metadata": {}, |
|
|
1110 |
"outputs": [ |
|
|
1111 |
{ |
|
|
1112 |
"data": { |
|
|
1113 |
"text/plain": [ |
|
|
1114 |
"array([85, 87, 88, 73, 75, 76, 63, 29], dtype=int32)" |
|
|
1115 |
] |
|
|
1116 |
}, |
|
|
1117 |
"execution_count": 54, |
|
|
1118 |
"metadata": {}, |
|
|
1119 |
"output_type": "execute_result" |
|
|
1120 |
} |
|
|
1121 |
], |
|
|
1122 |
"source": [ |
|
|
1123 |
"model = Hnsw(space='l2')\n", |
|
|
1124 |
"model.fit(coords)\n", |
|
|
1125 |
"model.query(coords, topn=8)" |
|
|
1126 |
] |
|
|
1127 |
}, |
|
|
1128 |
{ |
|
|
1129 |
"cell_type": "code", |
|
|
1130 |
"execution_count": 59, |
|
|
1131 |
"metadata": {}, |
|
|
1132 |
"outputs": [], |
|
|
1133 |
"source": [ |
|
|
1134 |
"import networkx as nx\n", |
|
|
1135 |
"G = nx.Graph()\n" |
|
|
1136 |
] |
|
|
1137 |
}, |
|
|
1138 |
{ |
|
|
1139 |
"cell_type": "code", |
|
|
1140 |
"execution_count": 56, |
|
|
1141 |
"metadata": {}, |
|
|
1142 |
"outputs": [ |
|
|
1143 |
{ |
|
|
1144 |
"data": { |
|
|
1145 |
"text/plain": [ |
|
|
1146 |
"array([43121, 29428])" |
|
|
1147 |
] |
|
|
1148 |
}, |
|
|
1149 |
"execution_count": 56, |
|
|
1150 |
"metadata": {}, |
|
|
1151 |
"output_type": "execute_result" |
|
|
1152 |
} |
|
|
1153 |
], |
|
|
1154 |
"source": [ |
|
|
1155 |
"for" |
|
|
1156 |
] |
|
|
1157 |
}, |
|
|
1158 |
{ |
|
|
1159 |
"cell_type": "code", |
|
|
1160 |
"execution_count": 52, |
|
|
1161 |
"metadata": {}, |
|
|
1162 |
"outputs": [ |
|
|
1163 |
{ |
|
|
1164 |
"data": { |
|
|
1165 |
"text/plain": [ |
|
|
1166 |
"130" |
|
|
1167 |
] |
|
|
1168 |
}, |
|
|
1169 |
"execution_count": 52, |
|
|
1170 |
"metadata": {}, |
|
|
1171 |
"output_type": "execute_result" |
|
|
1172 |
} |
|
|
1173 |
], |
|
|
1174 |
"source": [ |
|
|
1175 |
"temp[3]" |
|
|
1176 |
] |
|
|
1177 |
}, |
|
|
1178 |
{ |
|
|
1179 |
"cell_type": "code", |
|
|
1180 |
"execution_count": null, |
|
|
1181 |
"metadata": {}, |
|
|
1182 |
"outputs": [], |
|
|
1183 |
"source": [ |
|
|
1184 |
"model" |
|
|
1185 |
] |
|
|
1186 |
}, |
|
|
1187 |
{ |
|
|
1188 |
"cell_type": "code", |
|
|
1189 |
"execution_count": 29, |
|
|
1190 |
"metadata": {}, |
|
|
1191 |
"outputs": [ |
|
|
1192 |
{ |
|
|
1193 |
"data": { |
|
|
1194 |
"text/plain": [ |
|
|
1195 |
"array([ 7440, 13280])" |
|
|
1196 |
] |
|
|
1197 |
}, |
|
|
1198 |
"execution_count": 29, |
|
|
1199 |
"metadata": {}, |
|
|
1200 |
"output_type": "execute_result" |
|
|
1201 |
} |
|
|
1202 |
], |
|
|
1203 |
"source": [ |
|
|
1204 |
"coords[100]" |
|
|
1205 |
] |
|
|
1206 |
}, |
|
|
1207 |
{ |
|
|
1208 |
"cell_type": "code", |
|
|
1209 |
"execution_count": 33, |
|
|
1210 |
"metadata": {}, |
|
|
1211 |
"outputs": [], |
|
|
1212 |
"source": [ |
|
|
1213 |
"indices = model.query(coords[100], topn =10)" |
|
|
1214 |
] |
|
|
1215 |
}, |
|
|
1216 |
{ |
|
|
1217 |
"cell_type": "code", |
|
|
1218 |
"execution_count": 34, |
|
|
1219 |
"metadata": {}, |
|
|
1220 |
"outputs": [ |
|
|
1221 |
{ |
|
|
1222 |
"data": { |
|
|
1223 |
"text/plain": [ |
|
|
1224 |
"array([[ 7440, 13280],\n", |
|
|
1225 |
" [ 7440, 13792],\n", |
|
|
1226 |
" [ 7952, 13280],\n", |
|
|
1227 |
" [ 6928, 13792],\n", |
|
|
1228 |
" [ 7952, 12768],\n", |
|
|
1229 |
" [ 7952, 13792],\n", |
|
|
1230 |
" [ 7440, 14304],\n", |
|
|
1231 |
" [ 8464, 13280],\n", |
|
|
1232 |
" [ 6928, 14304],\n", |
|
|
1233 |
" [ 8464, 13792]])" |
|
|
1234 |
] |
|
|
1235 |
}, |
|
|
1236 |
"execution_count": 34, |
|
|
1237 |
"metadata": {}, |
|
|
1238 |
"output_type": "execute_result" |
|
|
1239 |
} |
|
|
1240 |
], |
|
|
1241 |
"source": [ |
|
|
1242 |
"coords[indices]" |
|
|
1243 |
] |
|
|
1244 |
}, |
|
|
1245 |
{ |
|
|
1246 |
"cell_type": "code", |
|
|
1247 |
"execution_count": 84, |
|
|
1248 |
"metadata": {}, |
|
|
1249 |
"outputs": [], |
|
|
1250 |
"source": [ |
|
|
1251 |
"def do_KmeansPCA(X=None, y=None, scaler=None, n_clusters=4, n_components=5):\n", |
|
|
1252 |
" import pandas as pd\n", |
|
|
1253 |
" import seaborn as sns\n", |
|
|
1254 |
" from sklearn.datasets import make_blobs\n", |
|
|
1255 |
" from sklearn import decomposition\n", |
|
|
1256 |
" from sklearn.decomposition import PCA, TruncatedSVD\n", |
|
|
1257 |
" from sklearn.preprocessing import StandardScaler, Normalizer\n", |
|
|
1258 |
" from sklearn.pipeline import make_pipeline\n", |
|
|
1259 |
" from sklearn.cluster import KMeans\n", |
|
|
1260 |
" ### Initialize Scaler\n", |
|
|
1261 |
" if scaler is None: \n", |
|
|
1262 |
" scaler = StandardScaler()\n", |
|
|
1263 |
" ### Get Random Data\n", |
|
|
1264 |
" X, y = make_blobs(n_features=10, n_samples=100, centers=4, random_state=4, cluster_std=7)\n", |
|
|
1265 |
" ### Scale Data\n", |
|
|
1266 |
" X = scaler.fit_transform(X)\n", |
|
|
1267 |
" ### Perform K-Means Clustering\n", |
|
|
1268 |
" cls = KMeans(n_clusters=n_clusters, init='k-means++', n_jobs=-1, n_init=1)\n", |
|
|
1269 |
" y_pred = cls.fit_predict(X)\n", |
|
|
1270 |
" ### Perform PCA\n", |
|
|
1271 |
" pca = PCA(n_components=n_components)\n", |
|
|
1272 |
" pc = pca.fit_transform(X)\n", |
|
|
1273 |
" ### Plot Results\n", |
|
|
1274 |
" columns = ['PC%d'%c for c in range(1, n_components+1)]\n", |
|
|
1275 |
" pc_df = pd.DataFrame(data=pc, columns=columns)\n", |
|
|
1276 |
" pc_df['y_pred'] = y_pred\n", |
|
|
1277 |
" pc_df['y'] = y\n", |
|
|
1278 |
" df = pd.DataFrame({'Variance Explained':pca.explained_variance_ratio_, 'Principal Components': columns})\n", |
|
|
1279 |
" sns.barplot(x='Principal Components',y=\"Variance Explained\", data=df, color=\"c\")\n", |
|
|
1280 |
" sns.lmplot( x=\"PC1\", y=\"PC2\", data=pc_df, fit_reg=False, \n", |
|
|
1281 |
" hue='y', legend=True, scatter_kws={\"s\": 80})\n", |
|
|
1282 |
" sns.lmplot( x=\"PC1\", y=\"PC2\", data=pc_df, fit_reg=False, \n", |
|
|
1283 |
" hue='y', legend=True, scatter_kws={\"s\": 80})" |
|
|
1284 |
] |
|
|
1285 |
}, |
|
|
1286 |
{ |
|
|
1287 |
"cell_type": "code", |
|
|
1288 |
"execution_count": 85, |
|
|
1289 |
"metadata": {}, |
|
|
1290 |
"outputs": [ |
|
|
1291 |
{ |
|
|
1292 |
"data": { |
|
|
1293 |
"image/png": "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\n", |
|
|
1294 |
"text/plain": [ |
|
|
1295 |
"<Figure size 432x288 with 1 Axes>" |
|
|
1296 |
] |
|
|
1297 |
}, |
|
|
1298 |
"metadata": { |
|
|
1299 |
"needs_background": "light" |
|
|
1300 |
}, |
|
|
1301 |
"output_type": "display_data" |
|
|
1302 |
}, |
|
|
1303 |
{ |
|
|
1304 |
"data": { |
|
|
1305 |
"image/png": "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\n", |
|
|
1306 |
"text/plain": [ |
|
|
1307 |
"<Figure size 402.375x360 with 1 Axes>" |
|
|
1308 |
] |
|
|
1309 |
}, |
|
|
1310 |
"metadata": { |
|
|
1311 |
"needs_background": "light" |
|
|
1312 |
}, |
|
|
1313 |
"output_type": "display_data" |
|
|
1314 |
}, |
|
|
1315 |
{ |
|
|
1316 |
"data": { |
|
|
1317 |
"image/png": "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\n", |
|
|
1318 |
"text/plain": [ |
|
|
1319 |
"<Figure size 402.375x360 with 1 Axes>" |
|
|
1320 |
] |
|
|
1321 |
}, |
|
|
1322 |
"metadata": { |
|
|
1323 |
"needs_background": "light" |
|
|
1324 |
}, |
|
|
1325 |
"output_type": "display_data" |
|
|
1326 |
} |
|
|
1327 |
], |
|
|
1328 |
"source": [ |
|
|
1329 |
"do_KmeansPCA()" |
|
|
1330 |
] |
|
|
1331 |
}, |
|
|
1332 |
{ |
|
|
1333 |
"cell_type": "code", |
|
|
1334 |
"execution_count": 76, |
|
|
1335 |
"metadata": {}, |
|
|
1336 |
"outputs": [ |
|
|
1337 |
{ |
|
|
1338 |
"data": { |
|
|
1339 |
"text/plain": [ |
|
|
1340 |
"(tensor([[1.0000, 0.0000, 1.0000, 0.9998]], grad_fn=<SigmoidBackward>),\n", |
|
|
1341 |
" tensor([[0., 0., 0., 0.]], grad_fn=<CumprodBackward>),\n", |
|
|
1342 |
" tensor([[2]]),\n", |
|
|
1343 |
" None,\n", |
|
|
1344 |
" None)" |
|
|
1345 |
] |
|
|
1346 |
}, |
|
|
1347 |
"execution_count": 76, |
|
|
1348 |
"metadata": {}, |
|
|
1349 |
"output_type": "execute_result" |
|
|
1350 |
} |
|
|
1351 |
], |
|
|
1352 |
"source": [ |
|
|
1353 |
"model.forward(x_path=x_path)" |
|
|
1354 |
] |
|
|
1355 |
}, |
|
|
1356 |
{ |
|
|
1357 |
"cell_type": "code", |
|
|
1358 |
"execution_count": 69, |
|
|
1359 |
"metadata": {}, |
|
|
1360 |
"outputs": [ |
|
|
1361 |
{ |
|
|
1362 |
"ename": "SyntaxError", |
|
|
1363 |
"evalue": "invalid syntax (<ipython-input-69-c543913fa78f>, line 1)", |
|
|
1364 |
"output_type": "error", |
|
|
1365 |
"traceback": [ |
|
|
1366 |
"\u001b[0;36m File \u001b[0;32m\"<ipython-input-69-c543913fa78f>\"\u001b[0;36m, line \u001b[0;32m1\u001b[0m\n\u001b[0;31m import ..models\u001b[0m\n\u001b[0m ^\u001b[0m\n\u001b[0;31mSyntaxError\u001b[0m\u001b[0;31m:\u001b[0m invalid syntax\n" |
|
|
1367 |
] |
|
|
1368 |
} |
|
|
1369 |
], |
|
|
1370 |
"source": [ |
|
|
1371 |
"import ..models" |
|
|
1372 |
] |
|
|
1373 |
}, |
|
|
1374 |
{ |
|
|
1375 |
"cell_type": "code", |
|
|
1376 |
"execution_count": 63, |
|
|
1377 |
"metadata": {}, |
|
|
1378 |
"outputs": [], |
|
|
1379 |
"source": [ |
|
|
1380 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.FloatTensor)" |
|
|
1381 |
] |
|
|
1382 |
}, |
|
|
1383 |
{ |
|
|
1384 |
"cell_type": "code", |
|
|
1385 |
"execution_count": 65, |
|
|
1386 |
"metadata": {}, |
|
|
1387 |
"outputs": [ |
|
|
1388 |
{ |
|
|
1389 |
"ename": "NameError", |
|
|
1390 |
"evalue": "name 'MultiheadAttention' is not defined", |
|
|
1391 |
"output_type": "error", |
|
|
1392 |
"traceback": [ |
|
|
1393 |
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", |
|
|
1394 |
"\u001b[0;31mNameError\u001b[0m Traceback (most recent call last)", |
|
|
1395 |
"\u001b[0;32m<ipython-input-65-f85a99af33ee>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mself\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mMM_CoAttn_Transformer_Surv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0momic_sizes\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0msig_sizes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", |
|
|
1396 |
"\u001b[0;32m<ipython-input-62-9e5f322e30a0>\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, omic_sizes, n_classes, model_size_wsi, model_size_omic, dropout)\u001b[0m\n\u001b[1;32m 28\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 29\u001b[0m \u001b[0;31m### Multihead Attention\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 30\u001b[0;31m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcoattn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mMultiheadAttention\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0membed_dim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m256\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_heads\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 31\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 32\u001b[0m \u001b[0;31m### Transformer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", |
|
|
1397 |
"\u001b[0;31mNameError\u001b[0m: name 'MultiheadAttention' is not defined" |
|
|
1398 |
] |
|
|
1399 |
} |
|
|
1400 |
], |
|
|
1401 |
"source": [ |
|
|
1402 |
"self = MM_CoAttn_Transformer_Surv(omic_sizes=sig_sizes)" |
|
|
1403 |
] |
|
|
1404 |
}, |
|
|
1405 |
{ |
|
|
1406 |
"cell_type": "code", |
|
|
1407 |
"execution_count": 52, |
|
|
1408 |
"metadata": {}, |
|
|
1409 |
"outputs": [ |
|
|
1410 |
{ |
|
|
1411 |
"ename": "NameError", |
|
|
1412 |
"evalue": "name 'sig_size' is not defined", |
|
|
1413 |
"output_type": "error", |
|
|
1414 |
"traceback": [ |
|
|
1415 |
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", |
|
|
1416 |
"\u001b[0;31mNameError\u001b[0m Traceback (most recent call last)", |
|
|
1417 |
"\u001b[0;32m<ipython-input-52-097a03ed0c40>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mself\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mMM_CoAttn_Surv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msig_sizes\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0msig_sizes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 2\u001b[0m \u001b[0mx_path\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m500\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1024\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mFloatTensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3\u001b[0m \u001b[0msig_feats\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mFloatTensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0msize\u001b[0m \u001b[0;32min\u001b[0m \u001b[0msig_sizes\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0mx_path\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mattention_net\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_path\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munsqueeze\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", |
|
|
1418 |
"\u001b[0;32m<ipython-input-43-4469ba9e1eea>\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, sig_sizes, n_classes, model_size_wsi, model_size_omic, dropout)\u001b[0m\n\u001b[1;32m 19\u001b[0m \u001b[0mhidden\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize_dict_omic\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mmodel_size_omic\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 20\u001b[0m \u001b[0msig_networks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 21\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0minput_dim\u001b[0m \u001b[0;32min\u001b[0m \u001b[0msig_size\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 22\u001b[0m \u001b[0mfc_omic\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mSNN_Block\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdim1\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minput_dim\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdim2\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mhidden\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 23\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m_\u001b[0m \u001b[0;32min\u001b[0m \u001b[0menumerate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhidden\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", |
|
|
1419 |
"\u001b[0;31mNameError\u001b[0m: name 'sig_size' is not defined" |
|
|
1420 |
] |
|
|
1421 |
} |
|
|
1422 |
], |
|
|
1423 |
"source": [ |
|
|
1424 |
"self = MM_CoAttn_Surv(sig_sizes=sig_sizes)\n", |
|
|
1425 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.FloatTensor)\n", |
|
|
1426 |
"sig_feats = [torch.randint(10, size=(size,)).type(torch.FloatTensor) for size in sig_sizes]\n", |
|
|
1427 |
"\n", |
|
|
1428 |
"x_path = self.attention_net(x_path).unsqueeze(1)\n", |
|
|
1429 |
"x_omic = torch.stack([self.sig_networks[idx].forward(sig_feat) for idx, sig_feat in enumerate(sig_feats)]).unsqueeze(1)\n", |
|
|
1430 |
"\n", |
|
|
1431 |
"out, attention_weights = self.coattn(x_omic, x_path, x_path)\n", |
|
|
1432 |
"out = self.transformer(out)\n", |
|
|
1433 |
"out = self.conv(out.squeeze(1).T.unsqueeze(0))\n", |
|
|
1434 |
"#out = self.classifier(out.squeeze(0).squeeze(1))" |
|
|
1435 |
] |
|
|
1436 |
}, |
|
|
1437 |
{ |
|
|
1438 |
"cell_type": "code", |
|
|
1439 |
"execution_count": 471, |
|
|
1440 |
"metadata": {}, |
|
|
1441 |
"outputs": [ |
|
|
1442 |
{ |
|
|
1443 |
"data": { |
|
|
1444 |
"text/plain": [ |
|
|
1445 |
"torch.Size([1, 256, 1])" |
|
|
1446 |
] |
|
|
1447 |
}, |
|
|
1448 |
"execution_count": 471, |
|
|
1449 |
"metadata": {}, |
|
|
1450 |
"output_type": "execute_result" |
|
|
1451 |
} |
|
|
1452 |
], |
|
|
1453 |
"source": [ |
|
|
1454 |
"out.shape" |
|
|
1455 |
] |
|
|
1456 |
}, |
|
|
1457 |
{ |
|
|
1458 |
"cell_type": "code", |
|
|
1459 |
"execution_count": 472, |
|
|
1460 |
"metadata": {}, |
|
|
1461 |
"outputs": [ |
|
|
1462 |
{ |
|
|
1463 |
"data": { |
|
|
1464 |
"text/plain": [ |
|
|
1465 |
"tensor([[[ 0.5998, 1.9873, -1.1435, ..., -0.0048, 0.2963, 1.1112]],\n", |
|
|
1466 |
"\n", |
|
|
1467 |
" [[-0.4201, -0.1456, 0.2057, ..., -0.2175, 0.4188, 0.4702]],\n", |
|
|
1468 |
"\n", |
|
|
1469 |
" [[ 1.0294, 3.1634, 0.4595, ..., 1.2059, 0.5845, 1.4114]],\n", |
|
|
1470 |
"\n", |
|
|
1471 |
" [[-1.1435, -1.1435, -1.1435, ..., 0.1951, -0.4378, 0.2051]],\n", |
|
|
1472 |
"\n", |
|
|
1473 |
" [[ 0.9948, 1.1596, 2.1419, ..., -0.1225, 1.3597, -0.3037]],\n", |
|
|
1474 |
"\n", |
|
|
1475 |
" [[ 0.4019, -1.1435, -0.1522, ..., -0.2058, 0.0351, -1.1435]]],\n", |
|
|
1476 |
" grad_fn=<UnsqueezeBackward0>)" |
|
|
1477 |
] |
|
|
1478 |
}, |
|
|
1479 |
"execution_count": 472, |
|
|
1480 |
"metadata": {}, |
|
|
1481 |
"output_type": "execute_result" |
|
|
1482 |
} |
|
|
1483 |
], |
|
|
1484 |
"source": [ |
|
|
1485 |
"x_omic" |
|
|
1486 |
] |
|
|
1487 |
}, |
|
|
1488 |
{ |
|
|
1489 |
"cell_type": "code", |
|
|
1490 |
"execution_count": 474, |
|
|
1491 |
"metadata": {}, |
|
|
1492 |
"outputs": [], |
|
|
1493 |
"source": [ |
|
|
1494 |
"self = MM_CoAttn_Surv(sig_sizes=sig_sizes)\n", |
|
|
1495 |
"x_path = torch.randint(10, size=(500, 1024)).type(torch.FloatTensor)\n", |
|
|
1496 |
"sig_feats = [torch.randint(10, size=(size,)).type(torch.FloatTensor) for size in sig_sizes]\n", |
|
|
1497 |
"\n", |
|
|
1498 |
"x_path = self.attention_net(x_path).unsqueeze(1)\n", |
|
|
1499 |
"x_omic = torch.stack([self.sig_networks[idx].forward(sig_feat) for idx, sig_feat in enumerate(sig_feats)]).unsqueeze(1)\n", |
|
|
1500 |
"out, attention_weights = self.coattn(x_omic, x_path, x_path)\n", |
|
|
1501 |
"\n", |
|
|
1502 |
"out = self.transformer(out)\n" |
|
|
1503 |
] |
|
|
1504 |
}, |
|
|
1505 |
{ |
|
|
1506 |
"cell_type": "code", |
|
|
1507 |
"execution_count": 491, |
|
|
1508 |
"metadata": {}, |
|
|
1509 |
"outputs": [ |
|
|
1510 |
{ |
|
|
1511 |
"data": { |
|
|
1512 |
"text/plain": [ |
|
|
1513 |
"torch.Size([1536])" |
|
|
1514 |
] |
|
|
1515 |
}, |
|
|
1516 |
"execution_count": 491, |
|
|
1517 |
"metadata": {}, |
|
|
1518 |
"output_type": "execute_result" |
|
|
1519 |
} |
|
|
1520 |
], |
|
|
1521 |
"source": [ |
|
|
1522 |
"torch.cat([self.sig_networks[idx].forward(sig_feat) for idx, sig_feat in enumerate(sig_feats)]).shape" |
|
|
1523 |
] |
|
|
1524 |
}, |
|
|
1525 |
{ |
|
|
1526 |
"cell_type": "code", |
|
|
1527 |
"execution_count": 484, |
|
|
1528 |
"metadata": {}, |
|
|
1529 |
"outputs": [ |
|
|
1530 |
{ |
|
|
1531 |
"data": { |
|
|
1532 |
"text/plain": [ |
|
|
1533 |
"torch.Size([6, 1, 512])" |
|
|
1534 |
] |
|
|
1535 |
}, |
|
|
1536 |
"execution_count": 484, |
|
|
1537 |
"metadata": {}, |
|
|
1538 |
"output_type": "execute_result" |
|
|
1539 |
} |
|
|
1540 |
], |
|
|
1541 |
"source": [ |
|
|
1542 |
"torch.cat([out, out], axis=2).shape" |
|
|
1543 |
] |
|
|
1544 |
}, |
|
|
1545 |
{ |
|
|
1546 |
"cell_type": "code", |
|
|
1547 |
"execution_count": 455, |
|
|
1548 |
"metadata": {}, |
|
|
1549 |
"outputs": [ |
|
|
1550 |
{ |
|
|
1551 |
"data": { |
|
|
1552 |
"text/plain": [ |
|
|
1553 |
"torch.Size([6, 1, 256])" |
|
|
1554 |
] |
|
|
1555 |
}, |
|
|
1556 |
"execution_count": 455, |
|
|
1557 |
"metadata": {}, |
|
|
1558 |
"output_type": "execute_result" |
|
|
1559 |
} |
|
|
1560 |
], |
|
|
1561 |
"source": [ |
|
|
1562 |
"out.shape" |
|
|
1563 |
] |
|
|
1564 |
}, |
|
|
1565 |
{ |
|
|
1566 |
"cell_type": "code", |
|
|
1567 |
"execution_count": 452, |
|
|
1568 |
"metadata": {}, |
|
|
1569 |
"outputs": [ |
|
|
1570 |
{ |
|
|
1571 |
"data": { |
|
|
1572 |
"text/plain": [ |
|
|
1573 |
"torch.Size([6, 1, 256])" |
|
|
1574 |
] |
|
|
1575 |
}, |
|
|
1576 |
"execution_count": 452, |
|
|
1577 |
"metadata": {}, |
|
|
1578 |
"output_type": "execute_result" |
|
|
1579 |
} |
|
|
1580 |
], |
|
|
1581 |
"source": [ |
|
|
1582 |
"out.shape" |
|
|
1583 |
] |
|
|
1584 |
}, |
|
|
1585 |
{ |
|
|
1586 |
"cell_type": "code", |
|
|
1587 |
"execution_count": null, |
|
|
1588 |
"metadata": {}, |
|
|
1589 |
"outputs": [], |
|
|
1590 |
"source": [] |
|
|
1591 |
}, |
|
|
1592 |
{ |
|
|
1593 |
"cell_type": "code", |
|
|
1594 |
"execution_count": 423, |
|
|
1595 |
"metadata": {}, |
|
|
1596 |
"outputs": [ |
|
|
1597 |
{ |
|
|
1598 |
"data": { |
|
|
1599 |
"text/plain": [ |
|
|
1600 |
"torch.Size([1, 8, 6, 500])" |
|
|
1601 |
] |
|
|
1602 |
}, |
|
|
1603 |
"execution_count": 423, |
|
|
1604 |
"metadata": {}, |
|
|
1605 |
"output_type": "execute_result" |
|
|
1606 |
} |
|
|
1607 |
], |
|
|
1608 |
"source": [ |
|
|
1609 |
"attention_weights.shape" |
|
|
1610 |
] |
|
|
1611 |
}, |
|
|
1612 |
{ |
|
|
1613 |
"cell_type": "code", |
|
|
1614 |
"execution_count": 415, |
|
|
1615 |
"metadata": {}, |
|
|
1616 |
"outputs": [ |
|
|
1617 |
{ |
|
|
1618 |
"data": { |
|
|
1619 |
"text/plain": [ |
|
|
1620 |
"tensor([[[0.0018, 0.0020, 0.0012, ..., 0.0016, 0.0025, 0.0031],\n", |
|
|
1621 |
" [0.0026, 0.0015, 0.0016, ..., 0.0021, 0.0021, 0.0016],\n", |
|
|
1622 |
" [0.0019, 0.0014, 0.0011, ..., 0.0020, 0.0013, 0.0025],\n", |
|
|
1623 |
" [0.0016, 0.0013, 0.0023, ..., 0.0009, 0.0015, 0.0027],\n", |
|
|
1624 |
" [0.0015, 0.0013, 0.0023, ..., 0.0026, 0.0019, 0.0026],\n", |
|
|
1625 |
" [0.0013, 0.0019, 0.0025, ..., 0.0022, 0.0020, 0.0021]]],\n", |
|
|
1626 |
" grad_fn=<DivBackward0>)" |
|
|
1627 |
] |
|
|
1628 |
}, |
|
|
1629 |
"execution_count": 415, |
|
|
1630 |
"metadata": {}, |
|
|
1631 |
"output_type": "execute_result" |
|
|
1632 |
} |
|
|
1633 |
], |
|
|
1634 |
"source": [ |
|
|
1635 |
"attention_weights_0" |
|
|
1636 |
] |
|
|
1637 |
}, |
|
|
1638 |
{ |
|
|
1639 |
"cell_type": "code", |
|
|
1640 |
"execution_count": 416, |
|
|
1641 |
"metadata": {}, |
|
|
1642 |
"outputs": [ |
|
|
1643 |
{ |
|
|
1644 |
"data": { |
|
|
1645 |
"text/plain": [ |
|
|
1646 |
"tensor([[[0.0018, 0.0020, 0.0012, ..., 0.0016, 0.0025, 0.0031],\n", |
|
|
1647 |
" [0.0026, 0.0015, 0.0016, ..., 0.0021, 0.0021, 0.0016],\n", |
|
|
1648 |
" [0.0019, 0.0014, 0.0011, ..., 0.0020, 0.0013, 0.0025],\n", |
|
|
1649 |
" [0.0016, 0.0013, 0.0023, ..., 0.0009, 0.0015, 0.0027],\n", |
|
|
1650 |
" [0.0015, 0.0013, 0.0023, ..., 0.0026, 0.0019, 0.0026],\n", |
|
|
1651 |
" [0.0013, 0.0019, 0.0025, ..., 0.0022, 0.0020, 0.0021]]],\n", |
|
|
1652 |
" grad_fn=<DivBackward0>)" |
|
|
1653 |
] |
|
|
1654 |
}, |
|
|
1655 |
"execution_count": 416, |
|
|
1656 |
"metadata": {}, |
|
|
1657 |
"output_type": "execute_result" |
|
|
1658 |
} |
|
|
1659 |
], |
|
|
1660 |
"source": [ |
|
|
1661 |
"softmax(attention_weights_1, dim=-1).sum(axis=1) / 8" |
|
|
1662 |
] |
|
|
1663 |
}, |
|
|
1664 |
{ |
|
|
1665 |
"cell_type": "code", |
|
|
1666 |
"execution_count": 411, |
|
|
1667 |
"metadata": {}, |
|
|
1668 |
"outputs": [ |
|
|
1669 |
{ |
|
|
1670 |
"data": { |
|
|
1671 |
"text/plain": [ |
|
|
1672 |
"torch.Size([1, 1, 6, 500])" |
|
|
1673 |
] |
|
|
1674 |
}, |
|
|
1675 |
"execution_count": 411, |
|
|
1676 |
"metadata": {}, |
|
|
1677 |
"output_type": "execute_result" |
|
|
1678 |
} |
|
|
1679 |
], |
|
|
1680 |
"source": [ |
|
|
1681 |
"softmax(attention_weights_1, dim=-1).shape" |
|
|
1682 |
] |
|
|
1683 |
}, |
|
|
1684 |
{ |
|
|
1685 |
"cell_type": "code", |
|
|
1686 |
"execution_count": 339, |
|
|
1687 |
"metadata": {}, |
|
|
1688 |
"outputs": [ |
|
|
1689 |
{ |
|
|
1690 |
"data": { |
|
|
1691 |
"text/plain": [ |
|
|
1692 |
"tensor(1.0000, grad_fn=<SumBackward0>)" |
|
|
1693 |
] |
|
|
1694 |
}, |
|
|
1695 |
"execution_count": 339, |
|
|
1696 |
"metadata": {}, |
|
|
1697 |
"output_type": "execute_result" |
|
|
1698 |
} |
|
|
1699 |
], |
|
|
1700 |
"source": [ |
|
|
1701 |
"attention_weights_0[0][0].sum()" |
|
|
1702 |
] |
|
|
1703 |
}, |
|
|
1704 |
{ |
|
|
1705 |
"cell_type": "code", |
|
|
1706 |
"execution_count": 396, |
|
|
1707 |
"metadata": {}, |
|
|
1708 |
"outputs": [], |
|
|
1709 |
"source": [ |
|
|
1710 |
"test = softmax(attention_weights_2, dim=-1)" |
|
|
1711 |
] |
|
|
1712 |
}, |
|
|
1713 |
{ |
|
|
1714 |
"cell_type": "code", |
|
|
1715 |
"execution_count": 402, |
|
|
1716 |
"metadata": {}, |
|
|
1717 |
"outputs": [ |
|
|
1718 |
{ |
|
|
1719 |
"data": { |
|
|
1720 |
"text/plain": [ |
|
|
1721 |
"tensor([0.0024, 0.0030, 0.0019, 0.0018, 0.0038, 0.0015, 0.0020, 0.0016, 0.0015,\n", |
|
|
1722 |
" 0.0019, 0.0015, 0.0035, 0.0026, 0.0017, 0.0014, 0.0013, 0.0023, 0.0020,\n", |
|
|
1723 |
" 0.0017, 0.0010], grad_fn=<SliceBackward>)" |
|
|
1724 |
] |
|
|
1725 |
}, |
|
|
1726 |
"execution_count": 402, |
|
|
1727 |
"metadata": {}, |
|
|
1728 |
"output_type": "execute_result" |
|
|
1729 |
} |
|
|
1730 |
], |
|
|
1731 |
"source": [ |
|
|
1732 |
"attention_weights_0[0][0][:20]" |
|
|
1733 |
] |
|
|
1734 |
}, |
|
|
1735 |
{ |
|
|
1736 |
"cell_type": "code", |
|
|
1737 |
"execution_count": 404, |
|
|
1738 |
"metadata": {}, |
|
|
1739 |
"outputs": [ |
|
|
1740 |
{ |
|
|
1741 |
"data": { |
|
|
1742 |
"text/plain": [ |
|
|
1743 |
"tensor([0.0028, 0.0033, 0.0019, 0.0013, 0.0042, 0.0016, 0.0024, 0.0018, 0.0019,\n", |
|
|
1744 |
" 0.0024, 0.0016, 0.0033, 0.0022, 0.0014, 0.0016, 0.0013, 0.0023, 0.0021,\n", |
|
|
1745 |
" 0.0013, 0.0013], grad_fn=<SliceBackward>)" |
|
|
1746 |
] |
|
|
1747 |
}, |
|
|
1748 |
"execution_count": 404, |
|
|
1749 |
"metadata": {}, |
|
|
1750 |
"output_type": "execute_result" |
|
|
1751 |
} |
|
|
1752 |
], |
|
|
1753 |
"source": [ |
|
|
1754 |
"test[0][0][:20]" |
|
|
1755 |
] |
|
|
1756 |
}, |
|
|
1757 |
{ |
|
|
1758 |
"cell_type": "code", |
|
|
1759 |
"execution_count": 366, |
|
|
1760 |
"metadata": {}, |
|
|
1761 |
"outputs": [ |
|
|
1762 |
{ |
|
|
1763 |
"data": { |
|
|
1764 |
"text/plain": [ |
|
|
1765 |
"tensor([[[False, False, False, ..., False, False, False],\n", |
|
|
1766 |
" [False, False, False, ..., False, False, False],\n", |
|
|
1767 |
" [False, False, False, ..., False, False, False],\n", |
|
|
1768 |
" [False, False, False, ..., False, False, False],\n", |
|
|
1769 |
" [False, False, False, ..., False, False, False],\n", |
|
|
1770 |
" [False, False, False, ..., False, False, False]]])" |
|
|
1771 |
] |
|
|
1772 |
}, |
|
|
1773 |
"execution_count": 366, |
|
|
1774 |
"metadata": {}, |
|
|
1775 |
"output_type": "execute_result" |
|
|
1776 |
} |
|
|
1777 |
], |
|
|
1778 |
"source": [ |
|
|
1779 |
"torch.eq(attention_weights_0, test)" |
|
|
1780 |
] |
|
|
1781 |
}, |
|
|
1782 |
{ |
|
|
1783 |
"cell_type": "code", |
|
|
1784 |
"execution_count": 320, |
|
|
1785 |
"metadata": {}, |
|
|
1786 |
"outputs": [ |
|
|
1787 |
{ |
|
|
1788 |
"data": { |
|
|
1789 |
"text/plain": [ |
|
|
1790 |
"torch.Size([1, 8, 6, 500])" |
|
|
1791 |
] |
|
|
1792 |
}, |
|
|
1793 |
"execution_count": 320, |
|
|
1794 |
"metadata": {}, |
|
|
1795 |
"output_type": "execute_result" |
|
|
1796 |
} |
|
|
1797 |
], |
|
|
1798 |
"source": [ |
|
|
1799 |
"attention_weights_1.shape" |
|
|
1800 |
] |
|
|
1801 |
}, |
|
|
1802 |
{ |
|
|
1803 |
"cell_type": "code", |
|
|
1804 |
"execution_count": 318, |
|
|
1805 |
"metadata": {}, |
|
|
1806 |
"outputs": [ |
|
|
1807 |
{ |
|
|
1808 |
"data": { |
|
|
1809 |
"text/plain": [ |
|
|
1810 |
"torch.Size([1, 6, 500])" |
|
|
1811 |
] |
|
|
1812 |
}, |
|
|
1813 |
"execution_count": 318, |
|
|
1814 |
"metadata": {}, |
|
|
1815 |
"output_type": "execute_result" |
|
|
1816 |
} |
|
|
1817 |
], |
|
|
1818 |
"source": [ |
|
|
1819 |
"attention_weights_2.shape" |
|
|
1820 |
] |
|
|
1821 |
}, |
|
|
1822 |
{ |
|
|
1823 |
"cell_type": "code", |
|
|
1824 |
"execution_count": 282, |
|
|
1825 |
"metadata": {}, |
|
|
1826 |
"outputs": [], |
|
|
1827 |
"source": [ |
|
|
1828 |
"out = self.classifier(out.squeeze(0).squeeze(1))" |
|
|
1829 |
] |
|
|
1830 |
}, |
|
|
1831 |
{ |
|
|
1832 |
"cell_type": "code", |
|
|
1833 |
"execution_count": 284, |
|
|
1834 |
"metadata": {}, |
|
|
1835 |
"outputs": [ |
|
|
1836 |
{ |
|
|
1837 |
"data": { |
|
|
1838 |
"text/plain": [ |
|
|
1839 |
"tensor([ 0.2832, 0.1548, -0.0972, -0.2801], grad_fn=<AddBackward0>)" |
|
|
1840 |
] |
|
|
1841 |
}, |
|
|
1842 |
"execution_count": 284, |
|
|
1843 |
"metadata": {}, |
|
|
1844 |
"output_type": "execute_result" |
|
|
1845 |
} |
|
|
1846 |
], |
|
|
1847 |
"source": [ |
|
|
1848 |
"out" |
|
|
1849 |
] |
|
|
1850 |
}, |
|
|
1851 |
{ |
|
|
1852 |
"cell_type": "code", |
|
|
1853 |
"execution_count": 269, |
|
|
1854 |
"metadata": {}, |
|
|
1855 |
"outputs": [ |
|
|
1856 |
{ |
|
|
1857 |
"data": { |
|
|
1858 |
"text/plain": [ |
|
|
1859 |
"tensor([[0.0018, 0.0019, 0.0019, ..., 0.0019, 0.0022, 0.0018],\n", |
|
|
1860 |
" [0.0020, 0.0020, 0.0021, ..., 0.0021, 0.0020, 0.0020],\n", |
|
|
1861 |
" [0.0019, 0.0022, 0.0021, ..., 0.0019, 0.0019, 0.0020],\n", |
|
|
1862 |
" [0.0021, 0.0022, 0.0019, ..., 0.0018, 0.0020, 0.0021],\n", |
|
|
1863 |
" [0.0019, 0.0019, 0.0020, ..., 0.0020, 0.0018, 0.0019],\n", |
|
|
1864 |
" [0.0021, 0.0021, 0.0019, ..., 0.0019, 0.0021, 0.0021]],\n", |
|
|
1865 |
" grad_fn=<SelectBackward>)" |
|
|
1866 |
] |
|
|
1867 |
}, |
|
|
1868 |
"execution_count": 269, |
|
|
1869 |
"metadata": {}, |
|
|
1870 |
"output_type": "execute_result" |
|
|
1871 |
} |
|
|
1872 |
], |
|
|
1873 |
"source": [ |
|
|
1874 |
"attention_weights[0]" |
|
|
1875 |
] |
|
|
1876 |
}, |
|
|
1877 |
{ |
|
|
1878 |
"cell_type": "code", |
|
|
1879 |
"execution_count": 241, |
|
|
1880 |
"metadata": {}, |
|
|
1881 |
"outputs": [ |
|
|
1882 |
{ |
|
|
1883 |
"data": { |
|
|
1884 |
"text/plain": [ |
|
|
1885 |
"(tensor([[[-0.0504, 0.0757, -0.0366, ..., -0.0275, -0.0294, 0.1300]],\n", |
|
|
1886 |
" \n", |
|
|
1887 |
" [[-0.0500, 0.0762, -0.0352, ..., -0.0253, -0.0289, 0.1311]],\n", |
|
|
1888 |
" \n", |
|
|
1889 |
" [[-0.0497, 0.0772, -0.0321, ..., -0.0246, -0.0288, 0.1301]],\n", |
|
|
1890 |
" \n", |
|
|
1891 |
" [[-0.0491, 0.0794, -0.0337, ..., -0.0260, -0.0278, 0.1281]],\n", |
|
|
1892 |
" \n", |
|
|
1893 |
" [[-0.0483, 0.0781, -0.0343, ..., -0.0246, -0.0301, 0.1321]],\n", |
|
|
1894 |
" \n", |
|
|
1895 |
" [[-0.0499, 0.0768, -0.0305, ..., -0.0257, -0.0280, 0.1321]]],\n", |
|
|
1896 |
" grad_fn=<AddBackward0>),\n", |
|
|
1897 |
" tensor([[[0.0019, 0.0019, 0.0019, ..., 0.0020, 0.0021, 0.0021],\n", |
|
|
1898 |
" [0.0017, 0.0020, 0.0020, ..., 0.0019, 0.0019, 0.0018],\n", |
|
|
1899 |
" [0.0019, 0.0018, 0.0019, ..., 0.0019, 0.0019, 0.0021],\n", |
|
|
1900 |
" [0.0020, 0.0020, 0.0019, ..., 0.0020, 0.0021, 0.0019],\n", |
|
|
1901 |
" [0.0017, 0.0023, 0.0021, ..., 0.0019, 0.0020, 0.0020],\n", |
|
|
1902 |
" [0.0021, 0.0021, 0.0020, ..., 0.0021, 0.0021, 0.0020]]],\n", |
|
|
1903 |
" grad_fn=<DivBackward0>))" |
|
|
1904 |
] |
|
|
1905 |
}, |
|
|
1906 |
"execution_count": 241, |
|
|
1907 |
"metadata": {}, |
|
|
1908 |
"output_type": "execute_result" |
|
|
1909 |
} |
|
|
1910 |
], |
|
|
1911 |
"source": [ |
|
|
1912 |
"self.coattn(x_omic, x_path, x_path)" |
|
|
1913 |
] |
|
|
1914 |
}, |
|
|
1915 |
{ |
|
|
1916 |
"cell_type": "code", |
|
|
1917 |
"execution_count": null, |
|
|
1918 |
"metadata": {}, |
|
|
1919 |
"outputs": [], |
|
|
1920 |
"source": [ |
|
|
1921 |
"h" |
|
|
1922 |
] |
|
|
1923 |
}, |
|
|
1924 |
{ |
|
|
1925 |
"cell_type": "code", |
|
|
1926 |
"execution_count": 208, |
|
|
1927 |
"metadata": {}, |
|
|
1928 |
"outputs": [], |
|
|
1929 |
"source": [ |
|
|
1930 |
"sig_feats = [torch.randn(size) for size in sig_sizes]\n", |
|
|
1931 |
"x_omic = torch.stack([self.sig_networks[idx].forward(sig_feat) for idx, sig_feat in enumerate(sig_feats)])\n" |
|
|
1932 |
] |
|
|
1933 |
}, |
|
|
1934 |
{ |
|
|
1935 |
"cell_type": "code", |
|
|
1936 |
"execution_count": 204, |
|
|
1937 |
"metadata": {}, |
|
|
1938 |
"outputs": [], |
|
|
1939 |
"source": [] |
|
|
1940 |
}, |
|
|
1941 |
{ |
|
|
1942 |
"cell_type": "code", |
|
|
1943 |
"execution_count": 206, |
|
|
1944 |
"metadata": {}, |
|
|
1945 |
"outputs": [ |
|
|
1946 |
{ |
|
|
1947 |
"data": { |
|
|
1948 |
"text/plain": [ |
|
|
1949 |
"torch.Size([6, 256])" |
|
|
1950 |
] |
|
|
1951 |
}, |
|
|
1952 |
"execution_count": 206, |
|
|
1953 |
"metadata": {}, |
|
|
1954 |
"output_type": "execute_result" |
|
|
1955 |
} |
|
|
1956 |
], |
|
|
1957 |
"source": [ |
|
|
1958 |
"x_omic.shape" |
|
|
1959 |
] |
|
|
1960 |
}, |
|
|
1961 |
{ |
|
|
1962 |
"cell_type": "code", |
|
|
1963 |
"execution_count": 166, |
|
|
1964 |
"metadata": {}, |
|
|
1965 |
"outputs": [ |
|
|
1966 |
{ |
|
|
1967 |
"ename": "NameError", |
|
|
1968 |
"evalue": "name 'sig1' is not defined", |
|
|
1969 |
"output_type": "error", |
|
|
1970 |
"traceback": [ |
|
|
1971 |
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", |
|
|
1972 |
"\u001b[0;31mNameError\u001b[0m Traceback (most recent call last)", |
|
|
1973 |
"\u001b[0;32m<ipython-input-166-aea4cb4c555c>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0msig1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msig2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msig3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msig4\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msig5\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msig6\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", |
|
|
1974 |
"\u001b[0;31mNameError\u001b[0m: name 'sig1' is not defined" |
|
|
1975 |
] |
|
|
1976 |
} |
|
|
1977 |
], |
|
|
1978 |
"source": [ |
|
|
1979 |
"sig1, sig2, sig3, sig4, sig5, sig6 = torch.randn()" |
|
|
1980 |
] |
|
|
1981 |
}, |
|
|
1982 |
{ |
|
|
1983 |
"cell_type": "code", |
|
|
1984 |
"execution_count": 158, |
|
|
1985 |
"metadata": {}, |
|
|
1986 |
"outputs": [], |
|
|
1987 |
"source": [ |
|
|
1988 |
"src = torch.rand(6, 1, 256)\n", |
|
|
1989 |
"out = transformer(src)\n", |
|
|
1990 |
"out = out.squeeze(1).T.unsqueeze(0)" |
|
|
1991 |
] |
|
|
1992 |
}, |
|
|
1993 |
{ |
|
|
1994 |
"cell_type": "code", |
|
|
1995 |
"execution_count": 163, |
|
|
1996 |
"metadata": {}, |
|
|
1997 |
"outputs": [], |
|
|
1998 |
"source": [ |
|
|
1999 |
"conv = nn.Conv1d(in_channels=256, out_channels=256, kernel_size=4, stride=4)" |
|
|
2000 |
] |
|
|
2001 |
}, |
|
|
2002 |
{ |
|
|
2003 |
"cell_type": "code", |
|
|
2004 |
"execution_count": 164, |
|
|
2005 |
"metadata": {}, |
|
|
2006 |
"outputs": [ |
|
|
2007 |
{ |
|
|
2008 |
"data": { |
|
|
2009 |
"text/plain": [ |
|
|
2010 |
"torch.Size([1, 256, 6])" |
|
|
2011 |
] |
|
|
2012 |
}, |
|
|
2013 |
"execution_count": 164, |
|
|
2014 |
"metadata": {}, |
|
|
2015 |
"output_type": "execute_result" |
|
|
2016 |
} |
|
|
2017 |
], |
|
|
2018 |
"source": [ |
|
|
2019 |
"out.shape" |
|
|
2020 |
] |
|
|
2021 |
}, |
|
|
2022 |
{ |
|
|
2023 |
"cell_type": "code", |
|
|
2024 |
"execution_count": 165, |
|
|
2025 |
"metadata": {}, |
|
|
2026 |
"outputs": [ |
|
|
2027 |
{ |
|
|
2028 |
"data": { |
|
|
2029 |
"text/plain": [ |
|
|
2030 |
"torch.Size([1, 256, 1])" |
|
|
2031 |
] |
|
|
2032 |
}, |
|
|
2033 |
"execution_count": 165, |
|
|
2034 |
"metadata": {}, |
|
|
2035 |
"output_type": "execute_result" |
|
|
2036 |
} |
|
|
2037 |
], |
|
|
2038 |
"source": [ |
|
|
2039 |
"conv(out).shape" |
|
|
2040 |
] |
|
|
2041 |
}, |
|
|
2042 |
{ |
|
|
2043 |
"cell_type": "code", |
|
|
2044 |
"execution_count": 112, |
|
|
2045 |
"metadata": {}, |
|
|
2046 |
"outputs": [ |
|
|
2047 |
{ |
|
|
2048 |
"data": { |
|
|
2049 |
"text/plain": [ |
|
|
2050 |
"torch.Size([1536])" |
|
|
2051 |
] |
|
|
2052 |
}, |
|
|
2053 |
"execution_count": 112, |
|
|
2054 |
"metadata": {}, |
|
|
2055 |
"output_type": "execute_result" |
|
|
2056 |
} |
|
|
2057 |
], |
|
|
2058 |
"source": [ |
|
|
2059 |
"x.reshape(-1).shape" |
|
|
2060 |
] |
|
|
2061 |
}, |
|
|
2062 |
{ |
|
|
2063 |
"cell_type": "code", |
|
|
2064 |
"execution_count": 106, |
|
|
2065 |
"metadata": {}, |
|
|
2066 |
"outputs": [ |
|
|
2067 |
{ |
|
|
2068 |
"data": { |
|
|
2069 |
"text/plain": [ |
|
|
2070 |
"3072" |
|
|
2071 |
] |
|
|
2072 |
}, |
|
|
2073 |
"execution_count": 106, |
|
|
2074 |
"metadata": {}, |
|
|
2075 |
"output_type": "execute_result" |
|
|
2076 |
} |
|
|
2077 |
], |
|
|
2078 |
"source": [ |
|
|
2079 |
"256 * 12" |
|
|
2080 |
] |
|
|
2081 |
}, |
|
|
2082 |
{ |
|
|
2083 |
"cell_type": "code", |
|
|
2084 |
"execution_count": 88, |
|
|
2085 |
"metadata": {}, |
|
|
2086 |
"outputs": [], |
|
|
2087 |
"source": [ |
|
|
2088 |
"net = Attn_Net_Gated()\n", |
|
|
2089 |
"wsi_feats = torch.randn(500, 1, 256)\n", |
|
|
2090 |
"sig_feats = torch.randn(6, 1, 256)" |
|
|
2091 |
] |
|
|
2092 |
}, |
|
|
2093 |
{ |
|
|
2094 |
"cell_type": "code", |
|
|
2095 |
"execution_count": 89, |
|
|
2096 |
"metadata": {}, |
|
|
2097 |
"outputs": [], |
|
|
2098 |
"source": [ |
|
|
2099 |
"multihead_attn = nn.MultiheadAttention(embed_dim=256, num_heads=8)" |
|
|
2100 |
] |
|
|
2101 |
}, |
|
|
2102 |
{ |
|
|
2103 |
"cell_type": "code", |
|
|
2104 |
"execution_count": 90, |
|
|
2105 |
"metadata": {}, |
|
|
2106 |
"outputs": [], |
|
|
2107 |
"source": [ |
|
|
2108 |
"out, coattn_weights = multihead_attn(sig_feats, wsi_feats, wsi_feats)" |
|
|
2109 |
] |
|
|
2110 |
}, |
|
|
2111 |
{ |
|
|
2112 |
"cell_type": "code", |
|
|
2113 |
"execution_count": 96, |
|
|
2114 |
"metadata": {}, |
|
|
2115 |
"outputs": [], |
|
|
2116 |
"source": [ |
|
|
2117 |
"cotton = DenseCoAttn(dim1=256, dim2=256, num_attn=8, num_none=3, dropout=0.3b)" |
|
|
2118 |
] |
|
|
2119 |
}, |
|
|
2120 |
{ |
|
|
2121 |
"cell_type": "code", |
|
|
2122 |
"execution_count": 100, |
|
|
2123 |
"metadata": {}, |
|
|
2124 |
"outputs": [], |
|
|
2125 |
"source": [ |
|
|
2126 |
"from math import sqrt\n", |
|
|
2127 |
"wsi_feats = torch.randn(1, 500, 256)\n", |
|
|
2128 |
"sig_feats = torch.randn(1, 6, 256)\n", |
|
|
2129 |
"_ = cotton(wsi_feats, sig_feats)" |
|
|
2130 |
] |
|
|
2131 |
}, |
|
|
2132 |
{ |
|
|
2133 |
"cell_type": "code", |
|
|
2134 |
"execution_count": 103, |
|
|
2135 |
"metadata": {}, |
|
|
2136 |
"outputs": [ |
|
|
2137 |
{ |
|
|
2138 |
"data": { |
|
|
2139 |
"text/plain": [ |
|
|
2140 |
"torch.Size([1, 6, 256])" |
|
|
2141 |
] |
|
|
2142 |
}, |
|
|
2143 |
"execution_count": 103, |
|
|
2144 |
"metadata": {}, |
|
|
2145 |
"output_type": "execute_result" |
|
|
2146 |
} |
|
|
2147 |
], |
|
|
2148 |
"source": [ |
|
|
2149 |
"_[0].shape" |
|
|
2150 |
] |
|
|
2151 |
}, |
|
|
2152 |
{ |
|
|
2153 |
"cell_type": "code", |
|
|
2154 |
"execution_count": 104, |
|
|
2155 |
"metadata": {}, |
|
|
2156 |
"outputs": [ |
|
|
2157 |
{ |
|
|
2158 |
"data": { |
|
|
2159 |
"text/plain": [ |
|
|
2160 |
"torch.Size([1, 500, 256])" |
|
|
2161 |
] |
|
|
2162 |
}, |
|
|
2163 |
"execution_count": 104, |
|
|
2164 |
"metadata": {}, |
|
|
2165 |
"output_type": "execute_result" |
|
|
2166 |
} |
|
|
2167 |
], |
|
|
2168 |
"source": [ |
|
|
2169 |
"_[1].shape" |
|
|
2170 |
] |
|
|
2171 |
}, |
|
|
2172 |
{ |
|
|
2173 |
"cell_type": "code", |
|
|
2174 |
"execution_count": 94, |
|
|
2175 |
"metadata": {}, |
|
|
2176 |
"outputs": [], |
|
|
2177 |
"source": [ |
|
|
2178 |
"\n", |
|
|
2179 |
"import torch\n", |
|
|
2180 |
"import torch.nn as nn\n", |
|
|
2181 |
"import torch.nn.functional as F\n", |
|
|
2182 |
"\n", |
|
|
2183 |
"\n", |
|
|
2184 |
"def qkv_attention(query, key, value, mask=None, dropout=None):\n", |
|
|
2185 |
"\td_k = query.size(-1)\n", |
|
|
2186 |
"\tscores = torch.matmul(query, key.transpose(-2,-1)) / sqrt(d_k)\n", |
|
|
2187 |
"\tif mask is not None:\n", |
|
|
2188 |
"\t\tscores.data.masked_fill_(mask.eq(0), -65504.0)\n", |
|
|
2189 |
"\t\n", |
|
|
2190 |
"\tp_attn = F.softmax(scores, dim=-1)\n", |
|
|
2191 |
"\tif dropout is not None:\n", |
|
|
2192 |
"\t\tp_attn = dropout(p_attn)\n", |
|
|
2193 |
"\n", |
|
|
2194 |
"\treturn torch.matmul(p_attn, value), p_attn\n", |
|
|
2195 |
"\n", |
|
|
2196 |
"\n", |
|
|
2197 |
"class DenseCoAttn(nn.Module):\n", |
|
|
2198 |
"\n", |
|
|
2199 |
"\tdef __init__(self, dim1, dim2, num_attn, num_none, dropout, is_multi_head=False):\n", |
|
|
2200 |
"\t\tsuper(DenseCoAttn, self).__init__()\n", |
|
|
2201 |
"\t\tdim = min(dim1, dim2)\n", |
|
|
2202 |
"\t\tself.linears = nn.ModuleList([nn.Linear(dim1, dim, bias=False),\n", |
|
|
2203 |
"\t\t\t\t\t\t\t\t\t nn.Linear(dim2, dim, bias=False)])\n", |
|
|
2204 |
"\t\tself.nones = nn.ParameterList([nn.Parameter(nn.init.xavier_uniform_(torch.empty(num_none, dim1))),\n", |
|
|
2205 |
"\t\t\t\t\t\t\t\t\t nn.Parameter(nn.init.xavier_uniform_(torch.empty(num_none, dim2)))])\n", |
|
|
2206 |
"\t\tself.d_k = dim // num_attn\n", |
|
|
2207 |
"\t\tself.h = num_attn\n", |
|
|
2208 |
"\t\tself.num_none = num_none\n", |
|
|
2209 |
"\t\tself.is_multi_head = is_multi_head\n", |
|
|
2210 |
"\t\tself.attn = None\n", |
|
|
2211 |
"\t\tself.dropouts = nn.ModuleList([nn.Dropout(p=dropout) for _ in range(2)])\n", |
|
|
2212 |
"\n", |
|
|
2213 |
"\tdef forward(self, value1, value2, mask1=None, mask2=None):\n", |
|
|
2214 |
"\t\tbatch = value1.size(0)\n", |
|
|
2215 |
"\t\tdim1, dim2 = value1.size(-1), value2.size(-1)\n", |
|
|
2216 |
"\t\tvalue1 = torch.cat([self.nones[0].unsqueeze(0).expand(batch, self.num_none, dim1), value1], dim=1)\n", |
|
|
2217 |
"\t\tvalue2 = torch.cat([self.nones[1].unsqueeze(0).expand(batch, self.num_none, dim2), value2], dim=1)\n", |
|
|
2218 |
"\t\tnone_mask = value1.new_ones((batch, self.num_none))\n", |
|
|
2219 |
"\n", |
|
|
2220 |
"\t\tif mask1 is not None:\n", |
|
|
2221 |
"\t\t\tmask1 = torch.cat([none_mask, mask1], dim=1)\n", |
|
|
2222 |
"\t\t\tmask1 = mask1.unsqueeze(1).unsqueeze(2)\n", |
|
|
2223 |
"\t\tif mask2 is not None:\n", |
|
|
2224 |
"\t\t\tmask2 = torch.cat([none_mask, mask2], dim=1)\n", |
|
|
2225 |
"\t\t\tmask2 = mask2.unsqueeze(1).unsqueeze(2)\n", |
|
|
2226 |
"\n", |
|
|
2227 |
"\t\tquery1, query2 = [l(x).view(batch, -1, self.h, self.d_k).transpose(1, 2) \n", |
|
|
2228 |
"\t\t\tfor l, x in zip(self.linears, (value1, value2))]\n", |
|
|
2229 |
"\n", |
|
|
2230 |
"\t\tif self.is_multi_head:\n", |
|
|
2231 |
"\t\t\tweighted1, attn1 = qkv_attention(query2, query1, query1, mask=mask1, dropout=self.dropouts[0])\n", |
|
|
2232 |
"\t\t\tweighted1 = weighted1.transpose(1, 2).contiguous()[:, self.num_none:, :]\n", |
|
|
2233 |
"\t\t\tweighted2, attn2 = qkv_attention(query1, query2, query2, mask=mask2, dropout=self.dropouts[1])\n", |
|
|
2234 |
"\t\t\tweighted2 = weighted2.transpose(1, 2).contiguous()[:, self.num_none:, :]\n", |
|
|
2235 |
"\t\telse:\n", |
|
|
2236 |
"\t\t\tweighted1, attn1 = qkv_attention(query2, query1, value1.unsqueeze(1), mask=mask1, \n", |
|
|
2237 |
"\t\t\t\tdropout=self.dropouts[0])\n", |
|
|
2238 |
"\t\t\tweighted1 = weighted1.mean(dim=1)[:, self.num_none:, :]\n", |
|
|
2239 |
"\t\t\tweighted2, attn2 = qkv_attention(query1, query2, value2.unsqueeze(1), mask=mask2, \n", |
|
|
2240 |
"\t\t\t\tdropout=self.dropouts[1])\n", |
|
|
2241 |
"\t\t\tweighted2 = weighted2.mean(dim=1)[:, self.num_none:, :]\n", |
|
|
2242 |
"\t\tself.attn = [attn1[:,:,self.num_none:,self.num_none:], attn2[:,:,self.num_none:,self.num_none:]]\n", |
|
|
2243 |
"\n", |
|
|
2244 |
"\t\treturn weighted1, weighted2\n" |
|
|
2245 |
] |
|
|
2246 |
}, |
|
|
2247 |
{ |
|
|
2248 |
"cell_type": "code", |
|
|
2249 |
"execution_count": null, |
|
|
2250 |
"metadata": {}, |
|
|
2251 |
"outputs": [], |
|
|
2252 |
"source": [] |
|
|
2253 |
}, |
|
|
2254 |
{ |
|
|
2255 |
"cell_type": "code", |
|
|
2256 |
"execution_count": null, |
|
|
2257 |
"metadata": {}, |
|
|
2258 |
"outputs": [], |
|
|
2259 |
"source": [] |
|
|
2260 |
}, |
|
|
2261 |
{ |
|
|
2262 |
"cell_type": "code", |
|
|
2263 |
"execution_count": null, |
|
|
2264 |
"metadata": {}, |
|
|
2265 |
"outputs": [], |
|
|
2266 |
"source": [] |
|
|
2267 |
}, |
|
|
2268 |
{ |
|
|
2269 |
"cell_type": "code", |
|
|
2270 |
"execution_count": 417, |
|
|
2271 |
"metadata": {}, |
|
|
2272 |
"outputs": [], |
|
|
2273 |
"source": [ |
|
|
2274 |
"from torch.nn.functional import *\n", |
|
|
2275 |
"\n", |
|
|
2276 |
"def multi_head_attention_forward(\n", |
|
|
2277 |
" query: Tensor,\n", |
|
|
2278 |
" key: Tensor,\n", |
|
|
2279 |
" value: Tensor,\n", |
|
|
2280 |
" embed_dim_to_check: int,\n", |
|
|
2281 |
" num_heads: int,\n", |
|
|
2282 |
" in_proj_weight: Tensor,\n", |
|
|
2283 |
" in_proj_bias: Tensor,\n", |
|
|
2284 |
" bias_k: Optional[Tensor],\n", |
|
|
2285 |
" bias_v: Optional[Tensor],\n", |
|
|
2286 |
" add_zero_attn: bool,\n", |
|
|
2287 |
" dropout_p: float,\n", |
|
|
2288 |
" out_proj_weight: Tensor,\n", |
|
|
2289 |
" out_proj_bias: Tensor,\n", |
|
|
2290 |
" training: bool = True,\n", |
|
|
2291 |
" key_padding_mask: Optional[Tensor] = None,\n", |
|
|
2292 |
" need_weights: bool = True,\n", |
|
|
2293 |
" need_raw: bool = True,\n", |
|
|
2294 |
" attn_mask: Optional[Tensor] = None,\n", |
|
|
2295 |
" use_separate_proj_weight: bool = False,\n", |
|
|
2296 |
" q_proj_weight: Optional[Tensor] = None,\n", |
|
|
2297 |
" k_proj_weight: Optional[Tensor] = None,\n", |
|
|
2298 |
" v_proj_weight: Optional[Tensor] = None,\n", |
|
|
2299 |
" static_k: Optional[Tensor] = None,\n", |
|
|
2300 |
" static_v: Optional[Tensor] = None,\n", |
|
|
2301 |
") -> Tuple[Tensor, Optional[Tensor]]:\n", |
|
|
2302 |
" r\"\"\"\n", |
|
|
2303 |
" Args:\n", |
|
|
2304 |
" query, key, value: map a query and a set of key-value pairs to an output.\n", |
|
|
2305 |
" See \"Attention Is All You Need\" for more details.\n", |
|
|
2306 |
" embed_dim_to_check: total dimension of the model.\n", |
|
|
2307 |
" num_heads: parallel attention heads.\n", |
|
|
2308 |
" in_proj_weight, in_proj_bias: input projection weight and bias.\n", |
|
|
2309 |
" bias_k, bias_v: bias of the key and value sequences to be added at dim=0.\n", |
|
|
2310 |
" add_zero_attn: add a new batch of zeros to the key and\n", |
|
|
2311 |
" value sequences at dim=1.\n", |
|
|
2312 |
" dropout_p: probability of an element to be zeroed.\n", |
|
|
2313 |
" out_proj_weight, out_proj_bias: the output projection weight and bias.\n", |
|
|
2314 |
" training: apply dropout if is ``True``.\n", |
|
|
2315 |
" key_padding_mask: if provided, specified padding elements in the key will\n", |
|
|
2316 |
" be ignored by the attention. This is an binary mask. When the value is True,\n", |
|
|
2317 |
" the corresponding value on the attention layer will be filled with -inf.\n", |
|
|
2318 |
" need_weights: output attn_output_weights.\n", |
|
|
2319 |
" attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all\n", |
|
|
2320 |
" the batches while a 3D mask allows to specify a different mask for the entries of each batch.\n", |
|
|
2321 |
" use_separate_proj_weight: the function accept the proj. weights for query, key,\n", |
|
|
2322 |
" and value in different forms. If false, in_proj_weight will be used, which is\n", |
|
|
2323 |
" a combination of q_proj_weight, k_proj_weight, v_proj_weight.\n", |
|
|
2324 |
" q_proj_weight, k_proj_weight, v_proj_weight, in_proj_bias: input projection weight and bias.\n", |
|
|
2325 |
" static_k, static_v: static key and value used for attention operators.\n", |
|
|
2326 |
" Shape:\n", |
|
|
2327 |
" Inputs:\n", |
|
|
2328 |
" - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is\n", |
|
|
2329 |
" the embedding dimension.\n", |
|
|
2330 |
" - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is\n", |
|
|
2331 |
" the embedding dimension.\n", |
|
|
2332 |
" - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is\n", |
|
|
2333 |
" the embedding dimension.\n", |
|
|
2334 |
" - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.\n", |
|
|
2335 |
" If a ByteTensor is provided, the non-zero positions will be ignored while the zero positions\n", |
|
|
2336 |
" will be unchanged. If a BoolTensor is provided, the positions with the\n", |
|
|
2337 |
" value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.\n", |
|
|
2338 |
" - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.\n", |
|
|
2339 |
" 3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,\n", |
|
|
2340 |
" S is the source sequence length. attn_mask ensures that position i is allowed to attend the unmasked\n", |
|
|
2341 |
" positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend\n", |
|
|
2342 |
" while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``\n", |
|
|
2343 |
" are not allowed to attend while ``False`` values will be unchanged. If a FloatTensor\n", |
|
|
2344 |
" is provided, it will be added to the attention weight.\n", |
|
|
2345 |
" - static_k: :math:`(N*num_heads, S, E/num_heads)`, where S is the source sequence length,\n", |
|
|
2346 |
" N is the batch size, E is the embedding dimension. E/num_heads is the head dimension.\n", |
|
|
2347 |
" - static_v: :math:`(N*num_heads, S, E/num_heads)`, where S is the source sequence length,\n", |
|
|
2348 |
" N is the batch size, E is the embedding dimension. E/num_heads is the head dimension.\n", |
|
|
2349 |
" Outputs:\n", |
|
|
2350 |
" - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,\n", |
|
|
2351 |
" E is the embedding dimension.\n", |
|
|
2352 |
" - attn_output_weights: :math:`(N, L, S)` where N is the batch size,\n", |
|
|
2353 |
" L is the target sequence length, S is the source sequence length.\n", |
|
|
2354 |
" \"\"\"\n", |
|
|
2355 |
" tens_ops = (query, key, value, in_proj_weight, in_proj_bias, bias_k, bias_v, out_proj_weight, out_proj_bias)\n", |
|
|
2356 |
" if has_torch_function(tens_ops):\n", |
|
|
2357 |
" return handle_torch_function(\n", |
|
|
2358 |
" multi_head_attention_forward,\n", |
|
|
2359 |
" tens_ops,\n", |
|
|
2360 |
" query,\n", |
|
|
2361 |
" key,\n", |
|
|
2362 |
" value,\n", |
|
|
2363 |
" embed_dim_to_check,\n", |
|
|
2364 |
" num_heads,\n", |
|
|
2365 |
" in_proj_weight,\n", |
|
|
2366 |
" in_proj_bias,\n", |
|
|
2367 |
" bias_k,\n", |
|
|
2368 |
" bias_v,\n", |
|
|
2369 |
" add_zero_attn,\n", |
|
|
2370 |
" dropout_p,\n", |
|
|
2371 |
" out_proj_weight,\n", |
|
|
2372 |
" out_proj_bias,\n", |
|
|
2373 |
" training=training,\n", |
|
|
2374 |
" key_padding_mask=key_padding_mask,\n", |
|
|
2375 |
" need_weights=need_weights,\n", |
|
|
2376 |
" need_raw=need_raw,\n", |
|
|
2377 |
" attn_mask=attn_mask,\n", |
|
|
2378 |
" use_separate_proj_weight=use_separate_proj_weight,\n", |
|
|
2379 |
" q_proj_weight=q_proj_weight,\n", |
|
|
2380 |
" k_proj_weight=k_proj_weight,\n", |
|
|
2381 |
" v_proj_weight=v_proj_weight,\n", |
|
|
2382 |
" static_k=static_k,\n", |
|
|
2383 |
" static_v=static_v,\n", |
|
|
2384 |
" )\n", |
|
|
2385 |
" tgt_len, bsz, embed_dim = query.size()\n", |
|
|
2386 |
" assert embed_dim == embed_dim_to_check\n", |
|
|
2387 |
" # allow MHA to have different sizes for the feature dimension\n", |
|
|
2388 |
" assert key.size(0) == value.size(0) and key.size(1) == value.size(1)\n", |
|
|
2389 |
"\n", |
|
|
2390 |
" head_dim = embed_dim // num_heads\n", |
|
|
2391 |
" assert head_dim * num_heads == embed_dim, \"embed_dim must be divisible by num_heads\"\n", |
|
|
2392 |
" scaling = float(head_dim) ** -0.5\n", |
|
|
2393 |
"\n", |
|
|
2394 |
" if not use_separate_proj_weight:\n", |
|
|
2395 |
" if (query is key or torch.equal(query, key)) and (key is value or torch.equal(key, value)):\n", |
|
|
2396 |
" # self-attention\n", |
|
|
2397 |
" q, k, v = linear(query, in_proj_weight, in_proj_bias).chunk(3, dim=-1)\n", |
|
|
2398 |
"\n", |
|
|
2399 |
" elif key is value or torch.equal(key, value):\n", |
|
|
2400 |
" # encoder-decoder attention\n", |
|
|
2401 |
" # This is inline in_proj function with in_proj_weight and in_proj_bias\n", |
|
|
2402 |
" _b = in_proj_bias\n", |
|
|
2403 |
" _start = 0\n", |
|
|
2404 |
" _end = embed_dim\n", |
|
|
2405 |
" _w = in_proj_weight[_start:_end, :]\n", |
|
|
2406 |
" if _b is not None:\n", |
|
|
2407 |
" _b = _b[_start:_end]\n", |
|
|
2408 |
" q = linear(query, _w, _b)\n", |
|
|
2409 |
"\n", |
|
|
2410 |
" if key is None:\n", |
|
|
2411 |
" assert value is None\n", |
|
|
2412 |
" k = None\n", |
|
|
2413 |
" v = None\n", |
|
|
2414 |
" else:\n", |
|
|
2415 |
"\n", |
|
|
2416 |
" # This is inline in_proj function with in_proj_weight and in_proj_bias\n", |
|
|
2417 |
" _b = in_proj_bias\n", |
|
|
2418 |
" _start = embed_dim\n", |
|
|
2419 |
" _end = None\n", |
|
|
2420 |
" _w = in_proj_weight[_start:, :]\n", |
|
|
2421 |
" if _b is not None:\n", |
|
|
2422 |
" _b = _b[_start:]\n", |
|
|
2423 |
" k, v = linear(key, _w, _b).chunk(2, dim=-1)\n", |
|
|
2424 |
"\n", |
|
|
2425 |
" else:\n", |
|
|
2426 |
" # This is inline in_proj function with in_proj_weight and in_proj_bias\n", |
|
|
2427 |
" _b = in_proj_bias\n", |
|
|
2428 |
" _start = 0\n", |
|
|
2429 |
" _end = embed_dim\n", |
|
|
2430 |
" _w = in_proj_weight[_start:_end, :]\n", |
|
|
2431 |
" if _b is not None:\n", |
|
|
2432 |
" _b = _b[_start:_end]\n", |
|
|
2433 |
" q = linear(query, _w, _b)\n", |
|
|
2434 |
"\n", |
|
|
2435 |
" # This is inline in_proj function with in_proj_weight and in_proj_bias\n", |
|
|
2436 |
" _b = in_proj_bias\n", |
|
|
2437 |
" _start = embed_dim\n", |
|
|
2438 |
" _end = embed_dim * 2\n", |
|
|
2439 |
" _w = in_proj_weight[_start:_end, :]\n", |
|
|
2440 |
" if _b is not None:\n", |
|
|
2441 |
" _b = _b[_start:_end]\n", |
|
|
2442 |
" k = linear(key, _w, _b)\n", |
|
|
2443 |
"\n", |
|
|
2444 |
" # This is inline in_proj function with in_proj_weight and in_proj_bias\n", |
|
|
2445 |
" _b = in_proj_bias\n", |
|
|
2446 |
" _start = embed_dim * 2\n", |
|
|
2447 |
" _end = None\n", |
|
|
2448 |
" _w = in_proj_weight[_start:, :]\n", |
|
|
2449 |
" if _b is not None:\n", |
|
|
2450 |
" _b = _b[_start:]\n", |
|
|
2451 |
" v = linear(value, _w, _b)\n", |
|
|
2452 |
" else:\n", |
|
|
2453 |
" q_proj_weight_non_opt = torch.jit._unwrap_optional(q_proj_weight)\n", |
|
|
2454 |
" len1, len2 = q_proj_weight_non_opt.size()\n", |
|
|
2455 |
" assert len1 == embed_dim and len2 == query.size(-1)\n", |
|
|
2456 |
"\n", |
|
|
2457 |
" k_proj_weight_non_opt = torch.jit._unwrap_optional(k_proj_weight)\n", |
|
|
2458 |
" len1, len2 = k_proj_weight_non_opt.size()\n", |
|
|
2459 |
" assert len1 == embed_dim and len2 == key.size(-1)\n", |
|
|
2460 |
"\n", |
|
|
2461 |
" v_proj_weight_non_opt = torch.jit._unwrap_optional(v_proj_weight)\n", |
|
|
2462 |
" len1, len2 = v_proj_weight_non_opt.size()\n", |
|
|
2463 |
" assert len1 == embed_dim and len2 == value.size(-1)\n", |
|
|
2464 |
"\n", |
|
|
2465 |
" if in_proj_bias is not None:\n", |
|
|
2466 |
" q = linear(query, q_proj_weight_non_opt, in_proj_bias[0:embed_dim])\n", |
|
|
2467 |
" k = linear(key, k_proj_weight_non_opt, in_proj_bias[embed_dim : (embed_dim * 2)])\n", |
|
|
2468 |
" v = linear(value, v_proj_weight_non_opt, in_proj_bias[(embed_dim * 2) :])\n", |
|
|
2469 |
" else:\n", |
|
|
2470 |
" q = linear(query, q_proj_weight_non_opt, in_proj_bias)\n", |
|
|
2471 |
" k = linear(key, k_proj_weight_non_opt, in_proj_bias)\n", |
|
|
2472 |
" v = linear(value, v_proj_weight_non_opt, in_proj_bias)\n", |
|
|
2473 |
" q = q * scaling\n", |
|
|
2474 |
"\n", |
|
|
2475 |
" if attn_mask is not None:\n", |
|
|
2476 |
" assert (\n", |
|
|
2477 |
" attn_mask.dtype == torch.float32\n", |
|
|
2478 |
" or attn_mask.dtype == torch.float64\n", |
|
|
2479 |
" or attn_mask.dtype == torch.float16\n", |
|
|
2480 |
" or attn_mask.dtype == torch.uint8\n", |
|
|
2481 |
" or attn_mask.dtype == torch.bool\n", |
|
|
2482 |
" ), \"Only float, byte, and bool types are supported for attn_mask, not {}\".format(attn_mask.dtype)\n", |
|
|
2483 |
" if attn_mask.dtype == torch.uint8:\n", |
|
|
2484 |
" warnings.warn(\"Byte tensor for attn_mask in nn.MultiheadAttention is deprecated. Use bool tensor instead.\")\n", |
|
|
2485 |
" attn_mask = attn_mask.to(torch.bool)\n", |
|
|
2486 |
"\n", |
|
|
2487 |
" if attn_mask.dim() == 2:\n", |
|
|
2488 |
" attn_mask = attn_mask.unsqueeze(0)\n", |
|
|
2489 |
" if list(attn_mask.size()) != [1, query.size(0), key.size(0)]:\n", |
|
|
2490 |
" raise RuntimeError(\"The size of the 2D attn_mask is not correct.\")\n", |
|
|
2491 |
" elif attn_mask.dim() == 3:\n", |
|
|
2492 |
" if list(attn_mask.size()) != [bsz * num_heads, query.size(0), key.size(0)]:\n", |
|
|
2493 |
" raise RuntimeError(\"The size of the 3D attn_mask is not correct.\")\n", |
|
|
2494 |
" else:\n", |
|
|
2495 |
" raise RuntimeError(\"attn_mask's dimension {} is not supported\".format(attn_mask.dim()))\n", |
|
|
2496 |
" # attn_mask's dim is 3 now.\n", |
|
|
2497 |
"\n", |
|
|
2498 |
" # convert ByteTensor key_padding_mask to bool\n", |
|
|
2499 |
" if key_padding_mask is not None and key_padding_mask.dtype == torch.uint8:\n", |
|
|
2500 |
" warnings.warn(\n", |
|
|
2501 |
" \"Byte tensor for key_padding_mask in nn.MultiheadAttention is deprecated. Use bool tensor instead.\"\n", |
|
|
2502 |
" )\n", |
|
|
2503 |
" key_padding_mask = key_padding_mask.to(torch.bool)\n", |
|
|
2504 |
"\n", |
|
|
2505 |
" if bias_k is not None and bias_v is not None:\n", |
|
|
2506 |
" if static_k is None and static_v is None:\n", |
|
|
2507 |
" k = torch.cat([k, bias_k.repeat(1, bsz, 1)])\n", |
|
|
2508 |
" v = torch.cat([v, bias_v.repeat(1, bsz, 1)])\n", |
|
|
2509 |
" if attn_mask is not None:\n", |
|
|
2510 |
" attn_mask = pad(attn_mask, (0, 1))\n", |
|
|
2511 |
" if key_padding_mask is not None:\n", |
|
|
2512 |
" key_padding_mask = pad(key_padding_mask, (0, 1))\n", |
|
|
2513 |
" else:\n", |
|
|
2514 |
" assert static_k is None, \"bias cannot be added to static key.\"\n", |
|
|
2515 |
" assert static_v is None, \"bias cannot be added to static value.\"\n", |
|
|
2516 |
" else:\n", |
|
|
2517 |
" assert bias_k is None\n", |
|
|
2518 |
" assert bias_v is None\n", |
|
|
2519 |
"\n", |
|
|
2520 |
" q = q.contiguous().view(tgt_len, bsz * num_heads, head_dim).transpose(0, 1)\n", |
|
|
2521 |
" if k is not None:\n", |
|
|
2522 |
" k = k.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)\n", |
|
|
2523 |
" if v is not None:\n", |
|
|
2524 |
" v = v.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)\n", |
|
|
2525 |
"\n", |
|
|
2526 |
" if static_k is not None:\n", |
|
|
2527 |
" assert static_k.size(0) == bsz * num_heads\n", |
|
|
2528 |
" assert static_k.size(2) == head_dim\n", |
|
|
2529 |
" k = static_k\n", |
|
|
2530 |
"\n", |
|
|
2531 |
" if static_v is not None:\n", |
|
|
2532 |
" assert static_v.size(0) == bsz * num_heads\n", |
|
|
2533 |
" assert static_v.size(2) == head_dim\n", |
|
|
2534 |
" v = static_v\n", |
|
|
2535 |
"\n", |
|
|
2536 |
" src_len = k.size(1)\n", |
|
|
2537 |
"\n", |
|
|
2538 |
" if key_padding_mask is not None:\n", |
|
|
2539 |
" assert key_padding_mask.size(0) == bsz\n", |
|
|
2540 |
" assert key_padding_mask.size(1) == src_len\n", |
|
|
2541 |
"\n", |
|
|
2542 |
" if add_zero_attn:\n", |
|
|
2543 |
" src_len += 1\n", |
|
|
2544 |
" k = torch.cat([k, torch.zeros((k.size(0), 1) + k.size()[2:], dtype=k.dtype, device=k.device)], dim=1)\n", |
|
|
2545 |
" v = torch.cat([v, torch.zeros((v.size(0), 1) + v.size()[2:], dtype=v.dtype, device=v.device)], dim=1)\n", |
|
|
2546 |
" if attn_mask is not None:\n", |
|
|
2547 |
" attn_mask = pad(attn_mask, (0, 1))\n", |
|
|
2548 |
" if key_padding_mask is not None:\n", |
|
|
2549 |
" key_padding_mask = pad(key_padding_mask, (0, 1))\n", |
|
|
2550 |
"\n", |
|
|
2551 |
" attn_output_weights = torch.bmm(q, k.transpose(1, 2))\n", |
|
|
2552 |
" assert list(attn_output_weights.size()) == [bsz * num_heads, tgt_len, src_len]\n", |
|
|
2553 |
"\n", |
|
|
2554 |
" if attn_mask is not None:\n", |
|
|
2555 |
" if attn_mask.dtype == torch.bool:\n", |
|
|
2556 |
" attn_output_weights.masked_fill_(attn_mask, float(\"-inf\"))\n", |
|
|
2557 |
" else:\n", |
|
|
2558 |
" attn_output_weights += attn_mask\n", |
|
|
2559 |
"\n", |
|
|
2560 |
" if key_padding_mask is not None:\n", |
|
|
2561 |
" attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len, src_len)\n", |
|
|
2562 |
" attn_output_weights = attn_output_weights.masked_fill(\n", |
|
|
2563 |
" key_padding_mask.unsqueeze(1).unsqueeze(2),\n", |
|
|
2564 |
" float(\"-inf\"),\n", |
|
|
2565 |
" )\n", |
|
|
2566 |
" attn_output_weights = attn_output_weights.view(bsz * num_heads, tgt_len, src_len)\n", |
|
|
2567 |
" \n", |
|
|
2568 |
" attn_output_weights_raw = attn_output_weights\n", |
|
|
2569 |
" attn_output_weights = softmax(attn_output_weights, dim=-1)\n", |
|
|
2570 |
" attn_output_weights = dropout(attn_output_weights, p=dropout_p, training=training)\n", |
|
|
2571 |
"\n", |
|
|
2572 |
" attn_output = torch.bmm(attn_output_weights, v)\n", |
|
|
2573 |
" assert list(attn_output.size()) == [bsz * num_heads, tgt_len, head_dim]\n", |
|
|
2574 |
" attn_output = attn_output.transpose(0, 1).contiguous().view(tgt_len, bsz, embed_dim)\n", |
|
|
2575 |
" attn_output = linear(attn_output, out_proj_weight, out_proj_bias)\n", |
|
|
2576 |
" \n", |
|
|
2577 |
" if need_weights:\n", |
|
|
2578 |
" if need_raw:\n", |
|
|
2579 |
" \n", |
|
|
2580 |
" attn_output_weights_raw = attn_output_weights_raw.view(bsz, num_heads, tgt_len, src_len)\n", |
|
|
2581 |
" return attn_output,attn_output_weights_raw\n", |
|
|
2582 |
" \n", |
|
|
2583 |
" #attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len, src_len)\n", |
|
|
2584 |
" #return attn_output, attn_output_weights.sum(dim=1) / num_heads, attn_output_weights_raw, attn_output_weights_raw.sum(dim=1) / num_heads\n", |
|
|
2585 |
" else:\n", |
|
|
2586 |
" # average attention weights over heads\n", |
|
|
2587 |
" attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len, src_len)\n", |
|
|
2588 |
" return attn_output, attn_output_weights.sum(dim=1) / num_heads\n", |
|
|
2589 |
" else:\n", |
|
|
2590 |
" return attn_output, None\n" |
|
|
2591 |
] |
|
|
2592 |
}, |
|
|
2593 |
{ |
|
|
2594 |
"cell_type": "code", |
|
|
2595 |
"execution_count": 418, |
|
|
2596 |
"metadata": {}, |
|
|
2597 |
"outputs": [], |
|
|
2598 |
"source": [ |
|
|
2599 |
"import torch\n", |
|
|
2600 |
"from torch import Tensor\n", |
|
|
2601 |
"from torch.nn.modules.linear import _LinearWithBias\n", |
|
|
2602 |
"from torch.nn.init import xavier_uniform_\n", |
|
|
2603 |
"from torch.nn.init import constant_\n", |
|
|
2604 |
"from torch.nn.init import xavier_normal_\n", |
|
|
2605 |
"from torch.nn.parameter import Parameter\n", |
|
|
2606 |
"from torch.nn import Module\n", |
|
|
2607 |
"\n", |
|
|
2608 |
"class MultiheadAttention(Module):\n", |
|
|
2609 |
" r\"\"\"Allows the model to jointly attend to information\n", |
|
|
2610 |
" from different representation subspaces.\n", |
|
|
2611 |
" See reference: Attention Is All You Need\n", |
|
|
2612 |
"\n", |
|
|
2613 |
" .. math::\n", |
|
|
2614 |
" \\text{MultiHead}(Q, K, V) = \\text{Concat}(head_1,\\dots,head_h)W^O\n", |
|
|
2615 |
" \\text{where} head_i = \\text{Attention}(QW_i^Q, KW_i^K, VW_i^V)\n", |
|
|
2616 |
"\n", |
|
|
2617 |
" Args:\n", |
|
|
2618 |
" embed_dim: total dimension of the model.\n", |
|
|
2619 |
" num_heads: parallel attention heads.\n", |
|
|
2620 |
" dropout: a Dropout layer on attn_output_weights. Default: 0.0.\n", |
|
|
2621 |
" bias: add bias as module parameter. Default: True.\n", |
|
|
2622 |
" add_bias_kv: add bias to the key and value sequences at dim=0.\n", |
|
|
2623 |
" add_zero_attn: add a new batch of zeros to the key and\n", |
|
|
2624 |
" value sequences at dim=1.\n", |
|
|
2625 |
" kdim: total number of features in key. Default: None.\n", |
|
|
2626 |
" vdim: total number of features in value. Default: None.\n", |
|
|
2627 |
"\n", |
|
|
2628 |
" Note: if kdim and vdim are None, they will be set to embed_dim such that\n", |
|
|
2629 |
" query, key, and value have the same number of features.\n", |
|
|
2630 |
"\n", |
|
|
2631 |
" Examples::\n", |
|
|
2632 |
"\n", |
|
|
2633 |
" >>> multihead_attn = nn.MultiheadAttention(embed_dim, num_heads)\n", |
|
|
2634 |
" >>> attn_output, attn_output_weights = multihead_attn(query, key, value)\n", |
|
|
2635 |
" \"\"\"\n", |
|
|
2636 |
" bias_k: Optional[torch.Tensor]\n", |
|
|
2637 |
" bias_v: Optional[torch.Tensor]\n", |
|
|
2638 |
"\n", |
|
|
2639 |
" def __init__(self, embed_dim, num_heads, dropout=0., bias=True, add_bias_kv=False, add_zero_attn=False, kdim=None, vdim=None):\n", |
|
|
2640 |
" super(MultiheadAttention, self).__init__()\n", |
|
|
2641 |
" self.embed_dim = embed_dim\n", |
|
|
2642 |
" self.kdim = kdim if kdim is not None else embed_dim\n", |
|
|
2643 |
" self.vdim = vdim if vdim is not None else embed_dim\n", |
|
|
2644 |
" self._qkv_same_embed_dim = self.kdim == embed_dim and self.vdim == embed_dim\n", |
|
|
2645 |
"\n", |
|
|
2646 |
" self.num_heads = num_heads\n", |
|
|
2647 |
" self.dropout = dropout\n", |
|
|
2648 |
" self.head_dim = embed_dim // num_heads\n", |
|
|
2649 |
" assert self.head_dim * num_heads == self.embed_dim, \"embed_dim must be divisible by num_heads\"\n", |
|
|
2650 |
"\n", |
|
|
2651 |
" if self._qkv_same_embed_dim is False:\n", |
|
|
2652 |
" self.q_proj_weight = Parameter(torch.Tensor(embed_dim, embed_dim))\n", |
|
|
2653 |
" self.k_proj_weight = Parameter(torch.Tensor(embed_dim, self.kdim))\n", |
|
|
2654 |
" self.v_proj_weight = Parameter(torch.Tensor(embed_dim, self.vdim))\n", |
|
|
2655 |
" self.register_parameter('in_proj_weight', None)\n", |
|
|
2656 |
" else:\n", |
|
|
2657 |
" self.in_proj_weight = Parameter(torch.empty(3 * embed_dim, embed_dim))\n", |
|
|
2658 |
" self.register_parameter('q_proj_weight', None)\n", |
|
|
2659 |
" self.register_parameter('k_proj_weight', None)\n", |
|
|
2660 |
" self.register_parameter('v_proj_weight', None)\n", |
|
|
2661 |
"\n", |
|
|
2662 |
" if bias:\n", |
|
|
2663 |
" self.in_proj_bias = Parameter(torch.empty(3 * embed_dim))\n", |
|
|
2664 |
" else:\n", |
|
|
2665 |
" self.register_parameter('in_proj_bias', None)\n", |
|
|
2666 |
" self.out_proj = _LinearWithBias(embed_dim, embed_dim)\n", |
|
|
2667 |
"\n", |
|
|
2668 |
" if add_bias_kv:\n", |
|
|
2669 |
" self.bias_k = Parameter(torch.empty(1, 1, embed_dim))\n", |
|
|
2670 |
" self.bias_v = Parameter(torch.empty(1, 1, embed_dim))\n", |
|
|
2671 |
" else:\n", |
|
|
2672 |
" self.bias_k = self.bias_v = None\n", |
|
|
2673 |
"\n", |
|
|
2674 |
" self.add_zero_attn = add_zero_attn\n", |
|
|
2675 |
"\n", |
|
|
2676 |
" self._reset_parameters()\n", |
|
|
2677 |
"\n", |
|
|
2678 |
" def _reset_parameters(self):\n", |
|
|
2679 |
" if self._qkv_same_embed_dim:\n", |
|
|
2680 |
" xavier_uniform_(self.in_proj_weight)\n", |
|
|
2681 |
" else:\n", |
|
|
2682 |
" xavier_uniform_(self.q_proj_weight)\n", |
|
|
2683 |
" xavier_uniform_(self.k_proj_weight)\n", |
|
|
2684 |
" xavier_uniform_(self.v_proj_weight)\n", |
|
|
2685 |
"\n", |
|
|
2686 |
" if self.in_proj_bias is not None:\n", |
|
|
2687 |
" constant_(self.in_proj_bias, 0.)\n", |
|
|
2688 |
" constant_(self.out_proj.bias, 0.)\n", |
|
|
2689 |
" if self.bias_k is not None:\n", |
|
|
2690 |
" xavier_normal_(self.bias_k)\n", |
|
|
2691 |
" if self.bias_v is not None:\n", |
|
|
2692 |
" xavier_normal_(self.bias_v)\n", |
|
|
2693 |
"\n", |
|
|
2694 |
" def __setstate__(self, state):\n", |
|
|
2695 |
" # Support loading old MultiheadAttention checkpoints generated by v1.1.0\n", |
|
|
2696 |
" if '_qkv_same_embed_dim' not in state:\n", |
|
|
2697 |
" state['_qkv_same_embed_dim'] = True\n", |
|
|
2698 |
"\n", |
|
|
2699 |
" super(MultiheadAttention, self).__setstate__(state)\n", |
|
|
2700 |
"\n", |
|
|
2701 |
" def forward(self, query, key, value, key_padding_mask=None,\n", |
|
|
2702 |
" need_weights=True, need_raw=True, attn_mask=None):\n", |
|
|
2703 |
" # type: (Tensor, Tensor, Tensor, Optional[Tensor], bool, Optional[Tensor]) -> Tuple[Tensor, Optional[Tensor]]\n", |
|
|
2704 |
" r\"\"\"\n", |
|
|
2705 |
" Args:\n", |
|
|
2706 |
" query, key, value: map a query and a set of key-value pairs to an output.\n", |
|
|
2707 |
" See \"Attention Is All You Need\" for more details.\n", |
|
|
2708 |
" key_padding_mask: if provided, specified padding elements in the key will\n", |
|
|
2709 |
" be ignored by the attention. When given a binary mask and a value is True,\n", |
|
|
2710 |
" the corresponding value on the attention layer will be ignored. When given\n", |
|
|
2711 |
" a byte mask and a value is non-zero, the corresponding value on the attention\n", |
|
|
2712 |
" layer will be ignored\n", |
|
|
2713 |
" need_weights: output attn_output_weights.\n", |
|
|
2714 |
" attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all\n", |
|
|
2715 |
" the batches while a 3D mask allows to specify a different mask for the entries of each batch.\n", |
|
|
2716 |
"\n", |
|
|
2717 |
" Shape:\n", |
|
|
2718 |
" - Inputs:\n", |
|
|
2719 |
" - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is\n", |
|
|
2720 |
" the embedding dimension.\n", |
|
|
2721 |
" - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is\n", |
|
|
2722 |
" the embedding dimension.\n", |
|
|
2723 |
" - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is\n", |
|
|
2724 |
" the embedding dimension.\n", |
|
|
2725 |
" - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.\n", |
|
|
2726 |
" If a ByteTensor is provided, the non-zero positions will be ignored while the position\n", |
|
|
2727 |
" with the zero positions will be unchanged. If a BoolTensor is provided, the positions with the\n", |
|
|
2728 |
" value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.\n", |
|
|
2729 |
" - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.\n", |
|
|
2730 |
" 3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,\n", |
|
|
2731 |
" S is the source sequence length. attn_mask ensure that position i is allowed to attend the unmasked\n", |
|
|
2732 |
" positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend\n", |
|
|
2733 |
" while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``\n", |
|
|
2734 |
" is not allowed to attend while ``False`` values will be unchanged. If a FloatTensor\n", |
|
|
2735 |
" is provided, it will be added to the attention weight.\n", |
|
|
2736 |
"\n", |
|
|
2737 |
" - Outputs:\n", |
|
|
2738 |
" - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,\n", |
|
|
2739 |
" E is the embedding dimension.\n", |
|
|
2740 |
" - attn_output_weights: :math:`(N, L, S)` where N is the batch size,\n", |
|
|
2741 |
" L is the target sequence length, S is the source sequence length.\n", |
|
|
2742 |
" \"\"\"\n", |
|
|
2743 |
" if not self._qkv_same_embed_dim:\n", |
|
|
2744 |
" return multi_head_attention_forward(\n", |
|
|
2745 |
" query, key, value, self.embed_dim, self.num_heads,\n", |
|
|
2746 |
" self.in_proj_weight, self.in_proj_bias,\n", |
|
|
2747 |
" self.bias_k, self.bias_v, self.add_zero_attn,\n", |
|
|
2748 |
" self.dropout, self.out_proj.weight, self.out_proj.bias,\n", |
|
|
2749 |
" training=self.training,\n", |
|
|
2750 |
" key_padding_mask=key_padding_mask, need_weights=need_weights, need_raw=need_raw,\n", |
|
|
2751 |
" attn_mask=attn_mask, use_separate_proj_weight=True,\n", |
|
|
2752 |
" q_proj_weight=self.q_proj_weight, k_proj_weight=self.k_proj_weight,\n", |
|
|
2753 |
" v_proj_weight=self.v_proj_weight)\n", |
|
|
2754 |
" else:\n", |
|
|
2755 |
" return multi_head_attention_forward(\n", |
|
|
2756 |
" query, key, value, self.embed_dim, self.num_heads,\n", |
|
|
2757 |
" self.in_proj_weight, self.in_proj_bias,\n", |
|
|
2758 |
" self.bias_k, self.bias_v, self.add_zero_attn,\n", |
|
|
2759 |
" self.dropout, self.out_proj.weight, self.out_proj.bias,\n", |
|
|
2760 |
" training=self.training,\n", |
|
|
2761 |
" key_padding_mask=key_padding_mask, need_weights=need_weights, need_raw=need_raw,\n", |
|
|
2762 |
" attn_mask=attn_mask)" |
|
|
2763 |
] |
|
|
2764 |
}, |
|
|
2765 |
{ |
|
|
2766 |
"cell_type": "code", |
|
|
2767 |
"execution_count": null, |
|
|
2768 |
"metadata": {}, |
|
|
2769 |
"outputs": [], |
|
|
2770 |
"source": [] |
|
|
2771 |
}, |
|
|
2772 |
{ |
|
|
2773 |
"cell_type": "code", |
|
|
2774 |
"execution_count": null, |
|
|
2775 |
"metadata": {}, |
|
|
2776 |
"outputs": [], |
|
|
2777 |
"source": [] |
|
|
2778 |
}, |
|
|
2779 |
{ |
|
|
2780 |
"cell_type": "code", |
|
|
2781 |
"execution_count": null, |
|
|
2782 |
"metadata": {}, |
|
|
2783 |
"outputs": [], |
|
|
2784 |
"source": [] |
|
|
2785 |
}, |
|
|
2786 |
{ |
|
|
2787 |
"cell_type": "code", |
|
|
2788 |
"execution_count": null, |
|
|
2789 |
"metadata": {}, |
|
|
2790 |
"outputs": [], |
|
|
2791 |
"source": [] |
|
|
2792 |
}, |
|
|
2793 |
{ |
|
|
2794 |
"cell_type": "code", |
|
|
2795 |
"execution_count": null, |
|
|
2796 |
"metadata": {}, |
|
|
2797 |
"outputs": [], |
|
|
2798 |
"source": [] |
|
|
2799 |
}, |
|
|
2800 |
{ |
|
|
2801 |
"cell_type": "code", |
|
|
2802 |
"execution_count": null, |
|
|
2803 |
"metadata": {}, |
|
|
2804 |
"outputs": [], |
|
|
2805 |
"source": [] |
|
|
2806 |
}, |
|
|
2807 |
{ |
|
|
2808 |
"cell_type": "code", |
|
|
2809 |
"execution_count": null, |
|
|
2810 |
"metadata": {}, |
|
|
2811 |
"outputs": [], |
|
|
2812 |
"source": [] |
|
|
2813 |
}, |
|
|
2814 |
{ |
|
|
2815 |
"cell_type": "code", |
|
|
2816 |
"execution_count": null, |
|
|
2817 |
"metadata": {}, |
|
|
2818 |
"outputs": [], |
|
|
2819 |
"source": [] |
|
|
2820 |
}, |
|
|
2821 |
{ |
|
|
2822 |
"cell_type": "code", |
|
|
2823 |
"execution_count": null, |
|
|
2824 |
"metadata": {}, |
|
|
2825 |
"outputs": [], |
|
|
2826 |
"source": [] |
|
|
2827 |
}, |
|
|
2828 |
{ |
|
|
2829 |
"cell_type": "code", |
|
|
2830 |
"execution_count": null, |
|
|
2831 |
"metadata": {}, |
|
|
2832 |
"outputs": [], |
|
|
2833 |
"source": [] |
|
|
2834 |
}, |
|
|
2835 |
{ |
|
|
2836 |
"cell_type": "code", |
|
|
2837 |
"execution_count": null, |
|
|
2838 |
"metadata": {}, |
|
|
2839 |
"outputs": [], |
|
|
2840 |
"source": [] |
|
|
2841 |
}, |
|
|
2842 |
{ |
|
|
2843 |
"cell_type": "code", |
|
|
2844 |
"execution_count": null, |
|
|
2845 |
"metadata": {}, |
|
|
2846 |
"outputs": [], |
|
|
2847 |
"source": [] |
|
|
2848 |
}, |
|
|
2849 |
{ |
|
|
2850 |
"cell_type": "code", |
|
|
2851 |
"execution_count": 104, |
|
|
2852 |
"metadata": {}, |
|
|
2853 |
"outputs": [ |
|
|
2854 |
{ |
|
|
2855 |
"ename": "ModuleNotFoundError", |
|
|
2856 |
"evalue": "No module named 'torch'", |
|
|
2857 |
"output_type": "error", |
|
|
2858 |
"traceback": [ |
|
|
2859 |
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", |
|
|
2860 |
"\u001b[0;31mModuleNotFoundError\u001b[0m Traceback (most recent call last)", |
|
|
2861 |
"\u001b[0;32m<ipython-input-104-6bb47b25d46a>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m 1\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mmath\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtorch\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", |
|
|
2862 |
"\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'torch'" |
|
|
2863 |
] |
|
|
2864 |
} |
|
|
2865 |
], |
|
|
2866 |
"source": [ |
|
|
2867 |
"import math\n", |
|
|
2868 |
"\n", |
|
|
2869 |
"import torch\n", |
|
|
2870 |
"from torch import nn\n", |
|
|
2871 |
"\n", |
|
|
2872 |
"############\n", |
|
|
2873 |
"# Omic Model\n", |
|
|
2874 |
"############\n", |
|
|
2875 |
"def init_max_weights(module):\n", |
|
|
2876 |
" for m in module.modules():\n", |
|
|
2877 |
" if type(m) == nn.Linear:\n", |
|
|
2878 |
" stdv = 1. / math.sqrt(m.weight.size(1))\n", |
|
|
2879 |
" m.weight.data.normal_(0, stdv)\n", |
|
|
2880 |
" m.bias.data.zero_()\n", |
|
|
2881 |
"\n", |
|
|
2882 |
"def SNN_Block(dim1, dim2, dropout=0.25):\n", |
|
|
2883 |
" return nn.Sequential(\n", |
|
|
2884 |
" nn.Linear(dim1, dim2),\n", |
|
|
2885 |
" nn.ELU(),\n", |
|
|
2886 |
" nn.AlphaDropout(p=dropout, inplace=False))\n", |
|
|
2887 |
"\n", |
|
|
2888 |
"class MaxNet(nn.Module):\n", |
|
|
2889 |
" def __init__(self, input_dim: int, meta_dim: int=0, model_size_omic: str='small', n_classes: int=4):\n", |
|
|
2890 |
" super(MaxNet, self).__init__()\n", |
|
|
2891 |
" self.meta_dim = meta_dim\n", |
|
|
2892 |
" self.n_classes = n_classes\n", |
|
|
2893 |
" self.size_dict_omic = {'small': [256, 256, 256, 256], 'big': [1024, 1024, 1024, 256]}\n", |
|
|
2894 |
" \n", |
|
|
2895 |
" ### Constructing Genomic SNN\n", |
|
|
2896 |
" hidden = self.size_dict_omic[model_size_omic]\n", |
|
|
2897 |
" fc_omic = [SNN_Block(dim1=input_dim, dim2=hidden[0])]\n", |
|
|
2898 |
" for i, _ in enumerate(hidden[1:]):\n", |
|
|
2899 |
" fc_omic.append(SNN_Block(dim1=hidden[i], dim2=hidden[i+1], dropout=0.25))\n", |
|
|
2900 |
" self.fc_omic = nn.Sequential(*fc_omic)\n", |
|
|
2901 |
" self.classifier = nn.Linear(hidden[-1]+self.meta_dim, n_classes)\n", |
|
|
2902 |
" init_max_weights(self)\n", |
|
|
2903 |
"\n", |
|
|
2904 |
" def forward(self, **kwargs):\n", |
|
|
2905 |
" x = kwargs['x_omic']\n", |
|
|
2906 |
" meta = kwargs['meta']\n", |
|
|
2907 |
" features = self.fc_omic(x)\n", |
|
|
2908 |
"\n", |
|
|
2909 |
" if self.meta_dim: \n", |
|
|
2910 |
" axis_dim = 1 if len(meta.shape) > 1 else 0\n", |
|
|
2911 |
" features = torch.cat((features, meta), axis_dim)\n", |
|
|
2912 |
"\n", |
|
|
2913 |
" logits = self.classifier(features).unsqueeze(0)\n", |
|
|
2914 |
" Y_hat = torch.topk(logits, 1, dim=1)[1]\n", |
|
|
2915 |
" hazards = torch.sigmoid(logits)\n", |
|
|
2916 |
" S = torch.cumprod(1 - hazards, dim=1)\n", |
|
|
2917 |
" return hazards, S, Y_hat, None, None\n", |
|
|
2918 |
"\n", |
|
|
2919 |
" def relocate(self):\n", |
|
|
2920 |
" device=torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n", |
|
|
2921 |
"\n", |
|
|
2922 |
" if torch.cuda.device_count() > 1:\n", |
|
|
2923 |
" device_ids = list(range(torch.cuda.device_count()))\n", |
|
|
2924 |
" self.fc_omic = nn.DataParallel(self.fc_omic, device_ids=device_ids).to('cuda:0')\n", |
|
|
2925 |
" else:\n", |
|
|
2926 |
" self.fc_omic = self.fc_omic.to(device)\n", |
|
|
2927 |
"\n", |
|
|
2928 |
"\n", |
|
|
2929 |
" self.classifier = self.classifier.to(device)" |
|
|
2930 |
] |
|
|
2931 |
}, |
|
|
2932 |
{ |
|
|
2933 |
"cell_type": "code", |
|
|
2934 |
"execution_count": null, |
|
|
2935 |
"metadata": {}, |
|
|
2936 |
"outputs": [], |
|
|
2937 |
"source": [] |
|
|
2938 |
}, |
|
|
2939 |
{ |
|
|
2940 |
"cell_type": "code", |
|
|
2941 |
"execution_count": 88, |
|
|
2942 |
"metadata": {}, |
|
|
2943 |
"outputs": [ |
|
|
2944 |
{ |
|
|
2945 |
"data": { |
|
|
2946 |
"text/html": [ |
|
|
2947 |
"<div>\n", |
|
|
2948 |
"<style scoped>\n", |
|
|
2949 |
" .dataframe tbody tr th:only-of-type {\n", |
|
|
2950 |
" vertical-align: middle;\n", |
|
|
2951 |
" }\n", |
|
|
2952 |
"\n", |
|
|
2953 |
" .dataframe tbody tr th {\n", |
|
|
2954 |
" vertical-align: top;\n", |
|
|
2955 |
" }\n", |
|
|
2956 |
"\n", |
|
|
2957 |
" .dataframe thead th {\n", |
|
|
2958 |
" text-align: right;\n", |
|
|
2959 |
" }\n", |
|
|
2960 |
"</style>\n", |
|
|
2961 |
"<table border=\"1\" class=\"dataframe\">\n", |
|
|
2962 |
" <thead>\n", |
|
|
2963 |
" <tr style=\"text-align: right;\">\n", |
|
|
2964 |
" <th></th>\n", |
|
|
2965 |
" <th>CXCL14_rnaseq</th>\n", |
|
|
2966 |
" <th>FGF1_rnaseq</th>\n", |
|
|
2967 |
" <th>IFNA8_cnv</th>\n", |
|
|
2968 |
" <th>ADM_rnaseq</th>\n", |
|
|
2969 |
" <th>LTBP2_rnaseq</th>\n", |
|
|
2970 |
" <th>CCL28_rnaseq</th>\n", |
|
|
2971 |
" <th>IFNA7_rnaseq</th>\n", |
|
|
2972 |
" <th>GH2_rnaseq</th>\n", |
|
|
2973 |
" <th>AIMP1_rnaseq</th>\n", |
|
|
2974 |
" <th>DEFB1_rnaseq</th>\n", |
|
|
2975 |
" <th>...</th>\n", |
|
|
2976 |
" <th>NPPB_rnaseq</th>\n", |
|
|
2977 |
" <th>CCL27_rnaseq</th>\n", |
|
|
2978 |
" <th>FASLG_rnaseq</th>\n", |
|
|
2979 |
" <th>FGF20_cnv</th>\n", |
|
|
2980 |
" <th>FAM3C_rnaseq</th>\n", |
|
|
2981 |
" <th>IL18_rnaseq</th>\n", |
|
|
2982 |
" <th>GDF10_rnaseq</th>\n", |
|
|
2983 |
" <th>MYDGF_rnaseq</th>\n", |
|
|
2984 |
" <th>IL10_rnaseq</th>\n", |
|
|
2985 |
" <th>IFNW1_rnaseq</th>\n", |
|
|
2986 |
" </tr>\n", |
|
|
2987 |
" </thead>\n", |
|
|
2988 |
" <tbody>\n", |
|
|
2989 |
" <tr>\n", |
|
|
2990 |
" <th>0</th>\n", |
|
|
2991 |
" <td>-0.1170</td>\n", |
|
|
2992 |
" <td>-0.2221</td>\n", |
|
|
2993 |
" <td>1</td>\n", |
|
|
2994 |
" <td>-0.5126</td>\n", |
|
|
2995 |
" <td>-0.3289</td>\n", |
|
|
2996 |
" <td>-0.7331</td>\n", |
|
|
2997 |
" <td>-0.1244</td>\n", |
|
|
2998 |
" <td>-0.1693</td>\n", |
|
|
2999 |
" <td>0.5942</td>\n", |
|
|
3000 |
" <td>-0.4707</td>\n", |
|
|
3001 |
" <td>...</td>\n", |
|
|
3002 |
" <td>-0.2276</td>\n", |
|
|
3003 |
" <td>1.2033</td>\n", |
|
|
3004 |
" <td>0.9826</td>\n", |
|
|
3005 |
" <td>-1</td>\n", |
|
|
3006 |
" <td>-0.6161</td>\n", |
|
|
3007 |
" <td>-0.5643</td>\n", |
|
|
3008 |
" <td>-0.2165</td>\n", |
|
|
3009 |
" <td>-0.2836</td>\n", |
|
|
3010 |
" <td>0.9991</td>\n", |
|
|
3011 |
" <td>-0.3899</td>\n", |
|
|
3012 |
" </tr>\n", |
|
|
3013 |
" <tr>\n", |
|
|
3014 |
" <th>1</th>\n", |
|
|
3015 |
" <td>-0.2330</td>\n", |
|
|
3016 |
" <td>-0.4343</td>\n", |
|
|
3017 |
" <td>-1</td>\n", |
|
|
3018 |
" <td>-0.2381</td>\n", |
|
|
3019 |
" <td>-0.4799</td>\n", |
|
|
3020 |
" <td>-0.0520</td>\n", |
|
|
3021 |
" <td>-0.1244</td>\n", |
|
|
3022 |
" <td>-0.1693</td>\n", |
|
|
3023 |
" <td>1.1854</td>\n", |
|
|
3024 |
" <td>-0.4820</td>\n", |
|
|
3025 |
" <td>...</td>\n", |
|
|
3026 |
" <td>-0.2276</td>\n", |
|
|
3027 |
" <td>-0.2946</td>\n", |
|
|
3028 |
" <td>-0.5443</td>\n", |
|
|
3029 |
" <td>-1</td>\n", |
|
|
3030 |
" <td>-0.3499</td>\n", |
|
|
3031 |
" <td>-0.7958</td>\n", |
|
|
3032 |
" <td>-0.3140</td>\n", |
|
|
3033 |
" <td>-0.3359</td>\n", |
|
|
3034 |
" <td>-0.4865</td>\n", |
|
|
3035 |
" <td>-0.3899</td>\n", |
|
|
3036 |
" </tr>\n", |
|
|
3037 |
" <tr>\n", |
|
|
3038 |
" <th>2</th>\n", |
|
|
3039 |
" <td>-0.1384</td>\n", |
|
|
3040 |
" <td>-0.1597</td>\n", |
|
|
3041 |
" <td>-1</td>\n", |
|
|
3042 |
" <td>-0.1521</td>\n", |
|
|
3043 |
" <td>-0.3348</td>\n", |
|
|
3044 |
" <td>-0.5310</td>\n", |
|
|
3045 |
" <td>-0.1244</td>\n", |
|
|
3046 |
" <td>-0.1693</td>\n", |
|
|
3047 |
" <td>0.3889</td>\n", |
|
|
3048 |
" <td>-0.3607</td>\n", |
|
|
3049 |
" <td>...</td>\n", |
|
|
3050 |
" <td>3.4177</td>\n", |
|
|
3051 |
" <td>-0.2946</td>\n", |
|
|
3052 |
" <td>-0.5320</td>\n", |
|
|
3053 |
" <td>0</td>\n", |
|
|
3054 |
" <td>0.4581</td>\n", |
|
|
3055 |
" <td>-0.6179</td>\n", |
|
|
3056 |
" <td>-0.2107</td>\n", |
|
|
3057 |
" <td>0.2751</td>\n", |
|
|
3058 |
" <td>-0.5108</td>\n", |
|
|
3059 |
" <td>1.0629</td>\n", |
|
|
3060 |
" </tr>\n", |
|
|
3061 |
" <tr>\n", |
|
|
3062 |
" <th>3</th>\n", |
|
|
3063 |
" <td>-0.1624</td>\n", |
|
|
3064 |
" <td>-0.3463</td>\n", |
|
|
3065 |
" <td>-1</td>\n", |
|
|
3066 |
" <td>0.0272</td>\n", |
|
|
3067 |
" <td>-0.7623</td>\n", |
|
|
3068 |
" <td>0.8196</td>\n", |
|
|
3069 |
" <td>-0.1244</td>\n", |
|
|
3070 |
" <td>-0.1693</td>\n", |
|
|
3071 |
" <td>-0.0416</td>\n", |
|
|
3072 |
" <td>0.1661</td>\n", |
|
|
3073 |
" <td>...</td>\n", |
|
|
3074 |
" <td>-0.2276</td>\n", |
|
|
3075 |
" <td>-0.1020</td>\n", |
|
|
3076 |
" <td>-0.4682</td>\n", |
|
|
3077 |
" <td>-1</td>\n", |
|
|
3078 |
" <td>-0.4391</td>\n", |
|
|
3079 |
" <td>-0.7275</td>\n", |
|
|
3080 |
" <td>-0.2876</td>\n", |
|
|
3081 |
" <td>-0.4696</td>\n", |
|
|
3082 |
" <td>-0.6248</td>\n", |
|
|
3083 |
" <td>-0.3899</td>\n", |
|
|
3084 |
" </tr>\n", |
|
|
3085 |
" <tr>\n", |
|
|
3086 |
" <th>4</th>\n", |
|
|
3087 |
" <td>-0.2346</td>\n", |
|
|
3088 |
" <td>-0.4090</td>\n", |
|
|
3089 |
" <td>-1</td>\n", |
|
|
3090 |
" <td>-0.2078</td>\n", |
|
|
3091 |
" <td>0.5702</td>\n", |
|
|
3092 |
" <td>-0.4219</td>\n", |
|
|
3093 |
" <td>-0.1244</td>\n", |
|
|
3094 |
" <td>0.5257</td>\n", |
|
|
3095 |
" <td>-0.9790</td>\n", |
|
|
3096 |
" <td>0.3938</td>\n", |
|
|
3097 |
" <td>...</td>\n", |
|
|
3098 |
" <td>-0.2276</td>\n", |
|
|
3099 |
" <td>-0.1035</td>\n", |
|
|
3100 |
" <td>-0.4688</td>\n", |
|
|
3101 |
" <td>-1</td>\n", |
|
|
3102 |
" <td>1.2596</td>\n", |
|
|
3103 |
" <td>-0.5807</td>\n", |
|
|
3104 |
" <td>0.4108</td>\n", |
|
|
3105 |
" <td>0.1801</td>\n", |
|
|
3106 |
" <td>-0.6086</td>\n", |
|
|
3107 |
" <td>-0.3899</td>\n", |
|
|
3108 |
" </tr>\n", |
|
|
3109 |
" <tr>\n", |
|
|
3110 |
" <th>...</th>\n", |
|
|
3111 |
" <td>...</td>\n", |
|
|
3112 |
" <td>...</td>\n", |
|
|
3113 |
" <td>...</td>\n", |
|
|
3114 |
" <td>...</td>\n", |
|
|
3115 |
" <td>...</td>\n", |
|
|
3116 |
" <td>...</td>\n", |
|
|
3117 |
" <td>...</td>\n", |
|
|
3118 |
" <td>...</td>\n", |
|
|
3119 |
" <td>...</td>\n", |
|
|
3120 |
" <td>...</td>\n", |
|
|
3121 |
" <td>...</td>\n", |
|
|
3122 |
" <td>...</td>\n", |
|
|
3123 |
" <td>...</td>\n", |
|
|
3124 |
" <td>...</td>\n", |
|
|
3125 |
" <td>...</td>\n", |
|
|
3126 |
" <td>...</td>\n", |
|
|
3127 |
" <td>...</td>\n", |
|
|
3128 |
" <td>...</td>\n", |
|
|
3129 |
" <td>...</td>\n", |
|
|
3130 |
" <td>...</td>\n", |
|
|
3131 |
" <td>...</td>\n", |
|
|
3132 |
" </tr>\n", |
|
|
3133 |
" <tr>\n", |
|
|
3134 |
" <th>368</th>\n", |
|
|
3135 |
" <td>-0.2417</td>\n", |
|
|
3136 |
" <td>10.1423</td>\n", |
|
|
3137 |
" <td>-1</td>\n", |
|
|
3138 |
" <td>-0.5456</td>\n", |
|
|
3139 |
" <td>0.8742</td>\n", |
|
|
3140 |
" <td>-0.1822</td>\n", |
|
|
3141 |
" <td>-0.1244</td>\n", |
|
|
3142 |
" <td>-0.1693</td>\n", |
|
|
3143 |
" <td>-1.2395</td>\n", |
|
|
3144 |
" <td>-0.5125</td>\n", |
|
|
3145 |
" <td>...</td>\n", |
|
|
3146 |
" <td>-0.2276</td>\n", |
|
|
3147 |
" <td>-0.2946</td>\n", |
|
|
3148 |
" <td>0.0777</td>\n", |
|
|
3149 |
" <td>0</td>\n", |
|
|
3150 |
" <td>-0.8242</td>\n", |
|
|
3151 |
" <td>-0.6727</td>\n", |
|
|
3152 |
" <td>0.1938</td>\n", |
|
|
3153 |
" <td>0.9210</td>\n", |
|
|
3154 |
" <td>0.4479</td>\n", |
|
|
3155 |
" <td>-0.3899</td>\n", |
|
|
3156 |
" </tr>\n", |
|
|
3157 |
" <tr>\n", |
|
|
3158 |
" <th>369</th>\n", |
|
|
3159 |
" <td>-0.2412</td>\n", |
|
|
3160 |
" <td>1.3253</td>\n", |
|
|
3161 |
" <td>1</td>\n", |
|
|
3162 |
" <td>-0.5680</td>\n", |
|
|
3163 |
" <td>1.0719</td>\n", |
|
|
3164 |
" <td>-0.1707</td>\n", |
|
|
3165 |
" <td>-0.1244</td>\n", |
|
|
3166 |
" <td>-0.1693</td>\n", |
|
|
3167 |
" <td>-1.6694</td>\n", |
|
|
3168 |
" <td>-0.4528</td>\n", |
|
|
3169 |
" <td>...</td>\n", |
|
|
3170 |
" <td>0.5679</td>\n", |
|
|
3171 |
" <td>-0.2661</td>\n", |
|
|
3172 |
" <td>1.0215</td>\n", |
|
|
3173 |
" <td>-2</td>\n", |
|
|
3174 |
" <td>-0.5327</td>\n", |
|
|
3175 |
" <td>0.3335</td>\n", |
|
|
3176 |
" <td>-0.1730</td>\n", |
|
|
3177 |
" <td>0.0147</td>\n", |
|
|
3178 |
" <td>0.6012</td>\n", |
|
|
3179 |
" <td>2.2526</td>\n", |
|
|
3180 |
" </tr>\n", |
|
|
3181 |
" <tr>\n", |
|
|
3182 |
" <th>370</th>\n", |
|
|
3183 |
" <td>-0.2396</td>\n", |
|
|
3184 |
" <td>0.0435</td>\n", |
|
|
3185 |
" <td>0</td>\n", |
|
|
3186 |
" <td>-0.3610</td>\n", |
|
|
3187 |
" <td>3.1965</td>\n", |
|
|
3188 |
" <td>1.3670</td>\n", |
|
|
3189 |
" <td>-0.1244</td>\n", |
|
|
3190 |
" <td>-0.1693</td>\n", |
|
|
3191 |
" <td>0.4439</td>\n", |
|
|
3192 |
" <td>-0.5099</td>\n", |
|
|
3193 |
" <td>...</td>\n", |
|
|
3194 |
" <td>-0.2276</td>\n", |
|
|
3195 |
" <td>-0.2289</td>\n", |
|
|
3196 |
" <td>0.0521</td>\n", |
|
|
3197 |
" <td>-1</td>\n", |
|
|
3198 |
" <td>1.0317</td>\n", |
|
|
3199 |
" <td>-0.1473</td>\n", |
|
|
3200 |
" <td>-0.1517</td>\n", |
|
|
3201 |
" <td>0.9384</td>\n", |
|
|
3202 |
" <td>-0.3165</td>\n", |
|
|
3203 |
" <td>0.6239</td>\n", |
|
|
3204 |
" </tr>\n", |
|
|
3205 |
" <tr>\n", |
|
|
3206 |
" <th>371</th>\n", |
|
|
3207 |
" <td>-0.2393</td>\n", |
|
|
3208 |
" <td>-0.4475</td>\n", |
|
|
3209 |
" <td>0</td>\n", |
|
|
3210 |
" <td>0.4772</td>\n", |
|
|
3211 |
" <td>2.9612</td>\n", |
|
|
3212 |
" <td>-0.7799</td>\n", |
|
|
3213 |
" <td>-0.1244</td>\n", |
|
|
3214 |
" <td>-0.1693</td>\n", |
|
|
3215 |
" <td>0.5778</td>\n", |
|
|
3216 |
" <td>1.7607</td>\n", |
|
|
3217 |
" <td>...</td>\n", |
|
|
3218 |
" <td>-0.2276</td>\n", |
|
|
3219 |
" <td>9.4098</td>\n", |
|
|
3220 |
" <td>-0.5443</td>\n", |
|
|
3221 |
" <td>0</td>\n", |
|
|
3222 |
" <td>0.2992</td>\n", |
|
|
3223 |
" <td>-0.5451</td>\n", |
|
|
3224 |
" <td>-0.2456</td>\n", |
|
|
3225 |
" <td>0.8898</td>\n", |
|
|
3226 |
" <td>-0.5781</td>\n", |
|
|
3227 |
" <td>-0.3899</td>\n", |
|
|
3228 |
" </tr>\n", |
|
|
3229 |
" <tr>\n", |
|
|
3230 |
" <th>372</th>\n", |
|
|
3231 |
" <td>-0.1936</td>\n", |
|
|
3232 |
" <td>-0.2281</td>\n", |
|
|
3233 |
" <td>0</td>\n", |
|
|
3234 |
" <td>-0.4124</td>\n", |
|
|
3235 |
" <td>-0.1873</td>\n", |
|
|
3236 |
" <td>-0.1200</td>\n", |
|
|
3237 |
" <td>-0.1244</td>\n", |
|
|
3238 |
" <td>-0.0326</td>\n", |
|
|
3239 |
" <td>-0.8786</td>\n", |
|
|
3240 |
" <td>-0.3912</td>\n", |
|
|
3241 |
" <td>...</td>\n", |
|
|
3242 |
" <td>-0.2276</td>\n", |
|
|
3243 |
" <td>-0.2570</td>\n", |
|
|
3244 |
" <td>-0.3810</td>\n", |
|
|
3245 |
" <td>-1</td>\n", |
|
|
3246 |
" <td>-0.6399</td>\n", |
|
|
3247 |
" <td>-0.9128</td>\n", |
|
|
3248 |
" <td>0.3367</td>\n", |
|
|
3249 |
" <td>-0.4686</td>\n", |
|
|
3250 |
" <td>0.8995</td>\n", |
|
|
3251 |
" <td>1.3522</td>\n", |
|
|
3252 |
" </tr>\n", |
|
|
3253 |
" </tbody>\n", |
|
|
3254 |
"</table>\n", |
|
|
3255 |
"<p>373 rows × 347 columns</p>\n", |
|
|
3256 |
"</div>" |
|
|
3257 |
], |
|
|
3258 |
"text/plain": [ |
|
|
3259 |
" CXCL14_rnaseq FGF1_rnaseq IFNA8_cnv ADM_rnaseq LTBP2_rnaseq \\\n", |
|
|
3260 |
"0 -0.1170 -0.2221 1 -0.5126 -0.3289 \n", |
|
|
3261 |
"1 -0.2330 -0.4343 -1 -0.2381 -0.4799 \n", |
|
|
3262 |
"2 -0.1384 -0.1597 -1 -0.1521 -0.3348 \n", |
|
|
3263 |
"3 -0.1624 -0.3463 -1 0.0272 -0.7623 \n", |
|
|
3264 |
"4 -0.2346 -0.4090 -1 -0.2078 0.5702 \n", |
|
|
3265 |
".. ... ... ... ... ... \n", |
|
|
3266 |
"368 -0.2417 10.1423 -1 -0.5456 0.8742 \n", |
|
|
3267 |
"369 -0.2412 1.3253 1 -0.5680 1.0719 \n", |
|
|
3268 |
"370 -0.2396 0.0435 0 -0.3610 3.1965 \n", |
|
|
3269 |
"371 -0.2393 -0.4475 0 0.4772 2.9612 \n", |
|
|
3270 |
"372 -0.1936 -0.2281 0 -0.4124 -0.1873 \n", |
|
|
3271 |
"\n", |
|
|
3272 |
" CCL28_rnaseq IFNA7_rnaseq GH2_rnaseq AIMP1_rnaseq DEFB1_rnaseq ... \\\n", |
|
|
3273 |
"0 -0.7331 -0.1244 -0.1693 0.5942 -0.4707 ... \n", |
|
|
3274 |
"1 -0.0520 -0.1244 -0.1693 1.1854 -0.4820 ... \n", |
|
|
3275 |
"2 -0.5310 -0.1244 -0.1693 0.3889 -0.3607 ... \n", |
|
|
3276 |
"3 0.8196 -0.1244 -0.1693 -0.0416 0.1661 ... \n", |
|
|
3277 |
"4 -0.4219 -0.1244 0.5257 -0.9790 0.3938 ... \n", |
|
|
3278 |
".. ... ... ... ... ... ... \n", |
|
|
3279 |
"368 -0.1822 -0.1244 -0.1693 -1.2395 -0.5125 ... \n", |
|
|
3280 |
"369 -0.1707 -0.1244 -0.1693 -1.6694 -0.4528 ... \n", |
|
|
3281 |
"370 1.3670 -0.1244 -0.1693 0.4439 -0.5099 ... \n", |
|
|
3282 |
"371 -0.7799 -0.1244 -0.1693 0.5778 1.7607 ... \n", |
|
|
3283 |
"372 -0.1200 -0.1244 -0.0326 -0.8786 -0.3912 ... \n", |
|
|
3284 |
"\n", |
|
|
3285 |
" NPPB_rnaseq CCL27_rnaseq FASLG_rnaseq FGF20_cnv FAM3C_rnaseq \\\n", |
|
|
3286 |
"0 -0.2276 1.2033 0.9826 -1 -0.6161 \n", |
|
|
3287 |
"1 -0.2276 -0.2946 -0.5443 -1 -0.3499 \n", |
|
|
3288 |
"2 3.4177 -0.2946 -0.5320 0 0.4581 \n", |
|
|
3289 |
"3 -0.2276 -0.1020 -0.4682 -1 -0.4391 \n", |
|
|
3290 |
"4 -0.2276 -0.1035 -0.4688 -1 1.2596 \n", |
|
|
3291 |
".. ... ... ... ... ... \n", |
|
|
3292 |
"368 -0.2276 -0.2946 0.0777 0 -0.8242 \n", |
|
|
3293 |
"369 0.5679 -0.2661 1.0215 -2 -0.5327 \n", |
|
|
3294 |
"370 -0.2276 -0.2289 0.0521 -1 1.0317 \n", |
|
|
3295 |
"371 -0.2276 9.4098 -0.5443 0 0.2992 \n", |
|
|
3296 |
"372 -0.2276 -0.2570 -0.3810 -1 -0.6399 \n", |
|
|
3297 |
"\n", |
|
|
3298 |
" IL18_rnaseq GDF10_rnaseq MYDGF_rnaseq IL10_rnaseq IFNW1_rnaseq \n", |
|
|
3299 |
"0 -0.5643 -0.2165 -0.2836 0.9991 -0.3899 \n", |
|
|
3300 |
"1 -0.7958 -0.3140 -0.3359 -0.4865 -0.3899 \n", |
|
|
3301 |
"2 -0.6179 -0.2107 0.2751 -0.5108 1.0629 \n", |
|
|
3302 |
"3 -0.7275 -0.2876 -0.4696 -0.6248 -0.3899 \n", |
|
|
3303 |
"4 -0.5807 0.4108 0.1801 -0.6086 -0.3899 \n", |
|
|
3304 |
".. ... ... ... ... ... \n", |
|
|
3305 |
"368 -0.6727 0.1938 0.9210 0.4479 -0.3899 \n", |
|
|
3306 |
"369 0.3335 -0.1730 0.0147 0.6012 2.2526 \n", |
|
|
3307 |
"370 -0.1473 -0.1517 0.9384 -0.3165 0.6239 \n", |
|
|
3308 |
"371 -0.5451 -0.2456 0.8898 -0.5781 -0.3899 \n", |
|
|
3309 |
"372 -0.9128 0.3367 -0.4686 0.8995 1.3522 \n", |
|
|
3310 |
"\n", |
|
|
3311 |
"[373 rows x 347 columns]" |
|
|
3312 |
] |
|
|
3313 |
}, |
|
|
3314 |
"execution_count": 88, |
|
|
3315 |
"metadata": {}, |
|
|
3316 |
"output_type": "execute_result" |
|
|
3317 |
} |
|
|
3318 |
], |
|
|
3319 |
"source": [ |
|
|
3320 |
"genomic_features[series_intersecdef series_intersection(s1, s2):\n", |
|
|
3321 |
" return pd.Series(list(set(s1) & set(s2)))\n", |
|
|
3322 |
"tion(sig, genomic_features.columns)]" |
|
|
3323 |
] |
|
|
3324 |
}, |
|
|
3325 |
{ |
|
|
3326 |
"cell_type": "code", |
|
|
3327 |
"execution_count": 84, |
|
|
3328 |
"metadata": {}, |
|
|
3329 |
"outputs": [], |
|
|
3330 |
"source": [ |
|
|
3331 |
"def series_intersection(s1, s2):\n", |
|
|
3332 |
" return pd.Series(list(set(s1) & set(s2)))\n" |
|
|
3333 |
] |
|
|
3334 |
}, |
|
|
3335 |
{ |
|
|
3336 |
"cell_type": "code", |
|
|
3337 |
"execution_count": 68, |
|
|
3338 |
"metadata": {}, |
|
|
3339 |
"outputs": [ |
|
|
3340 |
{ |
|
|
3341 |
"data": { |
|
|
3342 |
"text/html": [ |
|
|
3343 |
"<div>\n", |
|
|
3344 |
"<style scoped>\n", |
|
|
3345 |
" .dataframe tbody tr th:only-of-type {\n", |
|
|
3346 |
" vertical-align: middle;\n", |
|
|
3347 |
" }\n", |
|
|
3348 |
"\n", |
|
|
3349 |
" .dataframe tbody tr th {\n", |
|
|
3350 |
" vertical-align: top;\n", |
|
|
3351 |
" }\n", |
|
|
3352 |
"\n", |
|
|
3353 |
" .dataframe thead th {\n", |
|
|
3354 |
" text-align: right;\n", |
|
|
3355 |
" }\n", |
|
|
3356 |
"</style>\n", |
|
|
3357 |
"<table border=\"1\" class=\"dataframe\">\n", |
|
|
3358 |
" <thead>\n", |
|
|
3359 |
" <tr style=\"text-align: right;\">\n", |
|
|
3360 |
" <th></th>\n", |
|
|
3361 |
" <th>NDUFS5_cnv</th>\n", |
|
|
3362 |
" <th>MACF1_cnv</th>\n", |
|
|
3363 |
" <th>RNA5SP44_cnv</th>\n", |
|
|
3364 |
" <th>KIAA0754_cnv</th>\n", |
|
|
3365 |
" <th>BMP8A_cnv</th>\n", |
|
|
3366 |
" <th>PABPC4_cnv</th>\n", |
|
|
3367 |
" <th>SNORA55_cnv</th>\n", |
|
|
3368 |
" <th>HEYL_cnv</th>\n", |
|
|
3369 |
" <th>HPCAL4_cnv</th>\n", |
|
|
3370 |
" <th>NT5C1A_cnv</th>\n", |
|
|
3371 |
" <th>...</th>\n", |
|
|
3372 |
" <th>ZWINT_rnaseq</th>\n", |
|
|
3373 |
" <th>ZXDA_rnaseq</th>\n", |
|
|
3374 |
" <th>ZXDB_rnaseq</th>\n", |
|
|
3375 |
" <th>ZXDC_rnaseq</th>\n", |
|
|
3376 |
" <th>ZYG11A_rnaseq</th>\n", |
|
|
3377 |
" <th>ZYG11B_rnaseq</th>\n", |
|
|
3378 |
" <th>ZYX_rnaseq</th>\n", |
|
|
3379 |
" <th>ZZEF1_rnaseq</th>\n", |
|
|
3380 |
" <th>ZZZ3_rnaseq</th>\n", |
|
|
3381 |
" <th>TPTEP1_rnaseq</th>\n", |
|
|
3382 |
" </tr>\n", |
|
|
3383 |
" </thead>\n", |
|
|
3384 |
" <tbody>\n", |
|
|
3385 |
" <tr>\n", |
|
|
3386 |
" <th>0</th>\n", |
|
|
3387 |
" <td>-1</td>\n", |
|
|
3388 |
" <td>-1</td>\n", |
|
|
3389 |
" <td>-1</td>\n", |
|
|
3390 |
" <td>-1</td>\n", |
|
|
3391 |
" <td>-1</td>\n", |
|
|
3392 |
" <td>-1</td>\n", |
|
|
3393 |
" <td>-1</td>\n", |
|
|
3394 |
" <td>-1</td>\n", |
|
|
3395 |
" <td>-1</td>\n", |
|
|
3396 |
" <td>-1</td>\n", |
|
|
3397 |
" <td>...</td>\n", |
|
|
3398 |
" <td>-0.8388</td>\n", |
|
|
3399 |
" <td>4.1375</td>\n", |
|
|
3400 |
" <td>3.9664</td>\n", |
|
|
3401 |
" <td>1.8437</td>\n", |
|
|
3402 |
" <td>-0.3959</td>\n", |
|
|
3403 |
" <td>-0.2561</td>\n", |
|
|
3404 |
" <td>-0.2866</td>\n", |
|
|
3405 |
" <td>1.8770</td>\n", |
|
|
3406 |
" <td>-0.3179</td>\n", |
|
|
3407 |
" <td>-0.3633</td>\n", |
|
|
3408 |
" </tr>\n", |
|
|
3409 |
" <tr>\n", |
|
|
3410 |
" <th>1</th>\n", |
|
|
3411 |
" <td>2</td>\n", |
|
|
3412 |
" <td>2</td>\n", |
|
|
3413 |
" <td>2</td>\n", |
|
|
3414 |
" <td>2</td>\n", |
|
|
3415 |
" <td>2</td>\n", |
|
|
3416 |
" <td>2</td>\n", |
|
|
3417 |
" <td>2</td>\n", |
|
|
3418 |
" <td>2</td>\n", |
|
|
3419 |
" <td>2</td>\n", |
|
|
3420 |
" <td>2</td>\n", |
|
|
3421 |
" <td>...</td>\n", |
|
|
3422 |
" <td>-0.1083</td>\n", |
|
|
3423 |
" <td>0.3393</td>\n", |
|
|
3424 |
" <td>0.2769</td>\n", |
|
|
3425 |
" <td>1.7320</td>\n", |
|
|
3426 |
" <td>-0.0975</td>\n", |
|
|
3427 |
" <td>2.6955</td>\n", |
|
|
3428 |
" <td>-0.6741</td>\n", |
|
|
3429 |
" <td>1.0323</td>\n", |
|
|
3430 |
" <td>1.2766</td>\n", |
|
|
3431 |
" <td>-0.3982</td>\n", |
|
|
3432 |
" </tr>\n", |
|
|
3433 |
" <tr>\n", |
|
|
3434 |
" <th>2</th>\n", |
|
|
3435 |
" <td>0</td>\n", |
|
|
3436 |
" <td>0</td>\n", |
|
|
3437 |
" <td>0</td>\n", |
|
|
3438 |
" <td>0</td>\n", |
|
|
3439 |
" <td>0</td>\n", |
|
|
3440 |
" <td>0</td>\n", |
|
|
3441 |
" <td>0</td>\n", |
|
|
3442 |
" <td>0</td>\n", |
|
|
3443 |
" <td>0</td>\n", |
|
|
3444 |
" <td>0</td>\n", |
|
|
3445 |
" <td>...</td>\n", |
|
|
3446 |
" <td>-0.4155</td>\n", |
|
|
3447 |
" <td>1.6846</td>\n", |
|
|
3448 |
" <td>0.7711</td>\n", |
|
|
3449 |
" <td>-0.3061</td>\n", |
|
|
3450 |
" <td>-0.5016</td>\n", |
|
|
3451 |
" <td>2.8548</td>\n", |
|
|
3452 |
" <td>-0.6171</td>\n", |
|
|
3453 |
" <td>-0.8608</td>\n", |
|
|
3454 |
" <td>-0.0486</td>\n", |
|
|
3455 |
" <td>-0.3962</td>\n", |
|
|
3456 |
" </tr>\n", |
|
|
3457 |
" <tr>\n", |
|
|
3458 |
" <th>3</th>\n", |
|
|
3459 |
" <td>0</td>\n", |
|
|
3460 |
" <td>0</td>\n", |
|
|
3461 |
" <td>0</td>\n", |
|
|
3462 |
" <td>0</td>\n", |
|
|
3463 |
" <td>0</td>\n", |
|
|
3464 |
" <td>0</td>\n", |
|
|
3465 |
" <td>0</td>\n", |
|
|
3466 |
" <td>0</td>\n", |
|
|
3467 |
" <td>0</td>\n", |
|
|
3468 |
" <td>0</td>\n", |
|
|
3469 |
" <td>...</td>\n", |
|
|
3470 |
" <td>-0.8143</td>\n", |
|
|
3471 |
" <td>0.8344</td>\n", |
|
|
3472 |
" <td>1.5075</td>\n", |
|
|
3473 |
" <td>3.6068</td>\n", |
|
|
3474 |
" <td>-0.5004</td>\n", |
|
|
3475 |
" <td>-0.0747</td>\n", |
|
|
3476 |
" <td>-0.2185</td>\n", |
|
|
3477 |
" <td>-0.4379</td>\n", |
|
|
3478 |
" <td>1.6913</td>\n", |
|
|
3479 |
" <td>1.7748</td>\n", |
|
|
3480 |
" </tr>\n", |
|
|
3481 |
" <tr>\n", |
|
|
3482 |
" <th>4</th>\n", |
|
|
3483 |
" <td>0</td>\n", |
|
|
3484 |
" <td>0</td>\n", |
|
|
3485 |
" <td>0</td>\n", |
|
|
3486 |
" <td>0</td>\n", |
|
|
3487 |
" <td>0</td>\n", |
|
|
3488 |
" <td>0</td>\n", |
|
|
3489 |
" <td>0</td>\n", |
|
|
3490 |
" <td>0</td>\n", |
|
|
3491 |
" <td>0</td>\n", |
|
|
3492 |
" <td>0</td>\n", |
|
|
3493 |
" <td>...</td>\n", |
|
|
3494 |
" <td>0.0983</td>\n", |
|
|
3495 |
" <td>-0.7908</td>\n", |
|
|
3496 |
" <td>-0.0053</td>\n", |
|
|
3497 |
" <td>-0.0643</td>\n", |
|
|
3498 |
" <td>-0.3706</td>\n", |
|
|
3499 |
" <td>0.3870</td>\n", |
|
|
3500 |
" <td>-0.5589</td>\n", |
|
|
3501 |
" <td>-0.5979</td>\n", |
|
|
3502 |
" <td>0.0047</td>\n", |
|
|
3503 |
" <td>-0.3548</td>\n", |
|
|
3504 |
" </tr>\n", |
|
|
3505 |
" <tr>\n", |
|
|
3506 |
" <th>...</th>\n", |
|
|
3507 |
" <td>...</td>\n", |
|
|
3508 |
" <td>...</td>\n", |
|
|
3509 |
" <td>...</td>\n", |
|
|
3510 |
" <td>...</td>\n", |
|
|
3511 |
" <td>...</td>\n", |
|
|
3512 |
" <td>...</td>\n", |
|
|
3513 |
" <td>...</td>\n", |
|
|
3514 |
" <td>...</td>\n", |
|
|
3515 |
" <td>...</td>\n", |
|
|
3516 |
" <td>...</td>\n", |
|
|
3517 |
" <td>...</td>\n", |
|
|
3518 |
" <td>...</td>\n", |
|
|
3519 |
" <td>...</td>\n", |
|
|
3520 |
" <td>...</td>\n", |
|
|
3521 |
" <td>...</td>\n", |
|
|
3522 |
" <td>...</td>\n", |
|
|
3523 |
" <td>...</td>\n", |
|
|
3524 |
" <td>...</td>\n", |
|
|
3525 |
" <td>...</td>\n", |
|
|
3526 |
" <td>...</td>\n", |
|
|
3527 |
" <td>...</td>\n", |
|
|
3528 |
" </tr>\n", |
|
|
3529 |
" <tr>\n", |
|
|
3530 |
" <th>368</th>\n", |
|
|
3531 |
" <td>2</td>\n", |
|
|
3532 |
" <td>2</td>\n", |
|
|
3533 |
" <td>2</td>\n", |
|
|
3534 |
" <td>2</td>\n", |
|
|
3535 |
" <td>2</td>\n", |
|
|
3536 |
" <td>2</td>\n", |
|
|
3537 |
" <td>2</td>\n", |
|
|
3538 |
" <td>2</td>\n", |
|
|
3539 |
" <td>2</td>\n", |
|
|
3540 |
" <td>2</td>\n", |
|
|
3541 |
" <td>...</td>\n", |
|
|
3542 |
" <td>-0.0291</td>\n", |
|
|
3543 |
" <td>-0.1058</td>\n", |
|
|
3544 |
" <td>-0.6721</td>\n", |
|
|
3545 |
" <td>0.2802</td>\n", |
|
|
3546 |
" <td>1.9504</td>\n", |
|
|
3547 |
" <td>-0.8784</td>\n", |
|
|
3548 |
" <td>0.9506</td>\n", |
|
|
3549 |
" <td>0.0607</td>\n", |
|
|
3550 |
" <td>1.1883</td>\n", |
|
|
3551 |
" <td>-0.3521</td>\n", |
|
|
3552 |
" </tr>\n", |
|
|
3553 |
" <tr>\n", |
|
|
3554 |
" <th>369</th>\n", |
|
|
3555 |
" <td>0</td>\n", |
|
|
3556 |
" <td>0</td>\n", |
|
|
3557 |
" <td>0</td>\n", |
|
|
3558 |
" <td>0</td>\n", |
|
|
3559 |
" <td>0</td>\n", |
|
|
3560 |
" <td>0</td>\n", |
|
|
3561 |
" <td>0</td>\n", |
|
|
3562 |
" <td>0</td>\n", |
|
|
3563 |
" <td>0</td>\n", |
|
|
3564 |
" <td>0</td>\n", |
|
|
3565 |
" <td>...</td>\n", |
|
|
3566 |
" <td>0.0497</td>\n", |
|
|
3567 |
" <td>0.3673</td>\n", |
|
|
3568 |
" <td>-0.2208</td>\n", |
|
|
3569 |
" <td>0.3034</td>\n", |
|
|
3570 |
" <td>3.2580</td>\n", |
|
|
3571 |
" <td>-0.2089</td>\n", |
|
|
3572 |
" <td>1.6053</td>\n", |
|
|
3573 |
" <td>-0.8746</td>\n", |
|
|
3574 |
" <td>-0.4491</td>\n", |
|
|
3575 |
" <td>-0.3450</td>\n", |
|
|
3576 |
" </tr>\n", |
|
|
3577 |
" <tr>\n", |
|
|
3578 |
" <th>370</th>\n", |
|
|
3579 |
" <td>1</td>\n", |
|
|
3580 |
" <td>1</td>\n", |
|
|
3581 |
" <td>1</td>\n", |
|
|
3582 |
" <td>1</td>\n", |
|
|
3583 |
" <td>1</td>\n", |
|
|
3584 |
" <td>1</td>\n", |
|
|
3585 |
" <td>1</td>\n", |
|
|
3586 |
" <td>1</td>\n", |
|
|
3587 |
" <td>1</td>\n", |
|
|
3588 |
" <td>1</td>\n", |
|
|
3589 |
" <td>...</td>\n", |
|
|
3590 |
" <td>0.3822</td>\n", |
|
|
3591 |
" <td>-0.7003</td>\n", |
|
|
3592 |
" <td>-0.7661</td>\n", |
|
|
3593 |
" <td>-1.7035</td>\n", |
|
|
3594 |
" <td>-0.5423</td>\n", |
|
|
3595 |
" <td>-0.3488</td>\n", |
|
|
3596 |
" <td>1.3713</td>\n", |
|
|
3597 |
" <td>-0.4365</td>\n", |
|
|
3598 |
" <td>2.3456</td>\n", |
|
|
3599 |
" <td>-0.3866</td>\n", |
|
|
3600 |
" </tr>\n", |
|
|
3601 |
" <tr>\n", |
|
|
3602 |
" <th>371</th>\n", |
|
|
3603 |
" <td>0</td>\n", |
|
|
3604 |
" <td>0</td>\n", |
|
|
3605 |
" <td>0</td>\n", |
|
|
3606 |
" <td>0</td>\n", |
|
|
3607 |
" <td>0</td>\n", |
|
|
3608 |
" <td>0</td>\n", |
|
|
3609 |
" <td>0</td>\n", |
|
|
3610 |
" <td>0</td>\n", |
|
|
3611 |
" <td>0</td>\n", |
|
|
3612 |
" <td>0</td>\n", |
|
|
3613 |
" <td>...</td>\n", |
|
|
3614 |
" <td>-0.6853</td>\n", |
|
|
3615 |
" <td>-1.0240</td>\n", |
|
|
3616 |
" <td>-1.2890</td>\n", |
|
|
3617 |
" <td>-1.5666</td>\n", |
|
|
3618 |
" <td>-0.1270</td>\n", |
|
|
3619 |
" <td>-1.4662</td>\n", |
|
|
3620 |
" <td>0.3981</td>\n", |
|
|
3621 |
" <td>-0.5976</td>\n", |
|
|
3622 |
" <td>-1.3822</td>\n", |
|
|
3623 |
" <td>-0.4157</td>\n", |
|
|
3624 |
" </tr>\n", |
|
|
3625 |
" <tr>\n", |
|
|
3626 |
" <th>372</th>\n", |
|
|
3627 |
" <td>0</td>\n", |
|
|
3628 |
" <td>0</td>\n", |
|
|
3629 |
" <td>0</td>\n", |
|
|
3630 |
" <td>0</td>\n", |
|
|
3631 |
" <td>0</td>\n", |
|
|
3632 |
" <td>0</td>\n", |
|
|
3633 |
" <td>0</td>\n", |
|
|
3634 |
" <td>0</td>\n", |
|
|
3635 |
" <td>0</td>\n", |
|
|
3636 |
" <td>0</td>\n", |
|
|
3637 |
" <td>...</td>\n", |
|
|
3638 |
" <td>0.0517</td>\n", |
|
|
3639 |
" <td>-0.3570</td>\n", |
|
|
3640 |
" <td>-0.4843</td>\n", |
|
|
3641 |
" <td>-0.3792</td>\n", |
|
|
3642 |
" <td>-0.1964</td>\n", |
|
|
3643 |
" <td>0.4200</td>\n", |
|
|
3644 |
" <td>3.2547</td>\n", |
|
|
3645 |
" <td>-0.1232</td>\n", |
|
|
3646 |
" <td>3.4519</td>\n", |
|
|
3647 |
" <td>-0.1962</td>\n", |
|
|
3648 |
" </tr>\n", |
|
|
3649 |
" </tbody>\n", |
|
|
3650 |
"</table>\n", |
|
|
3651 |
"<p>373 rows × 20395 columns</p>\n", |
|
|
3652 |
"</div>" |
|
|
3653 |
], |
|
|
3654 |
"text/plain": [ |
|
|
3655 |
" NDUFS5_cnv MACF1_cnv RNA5SP44_cnv KIAA0754_cnv BMP8A_cnv PABPC4_cnv \\\n", |
|
|
3656 |
"0 -1 -1 -1 -1 -1 -1 \n", |
|
|
3657 |
"1 2 2 2 2 2 2 \n", |
|
|
3658 |
"2 0 0 0 0 0 0 \n", |
|
|
3659 |
"3 0 0 0 0 0 0 \n", |
|
|
3660 |
"4 0 0 0 0 0 0 \n", |
|
|
3661 |
".. ... ... ... ... ... ... \n", |
|
|
3662 |
"368 2 2 2 2 2 2 \n", |
|
|
3663 |
"369 0 0 0 0 0 0 \n", |
|
|
3664 |
"370 1 1 1 1 1 1 \n", |
|
|
3665 |
"371 0 0 0 0 0 0 \n", |
|
|
3666 |
"372 0 0 0 0 0 0 \n", |
|
|
3667 |
"\n", |
|
|
3668 |
" SNORA55_cnv HEYL_cnv HPCAL4_cnv NT5C1A_cnv ... ZWINT_rnaseq \\\n", |
|
|
3669 |
"0 -1 -1 -1 -1 ... -0.8388 \n", |
|
|
3670 |
"1 2 2 2 2 ... -0.1083 \n", |
|
|
3671 |
"2 0 0 0 0 ... -0.4155 \n", |
|
|
3672 |
"3 0 0 0 0 ... -0.8143 \n", |
|
|
3673 |
"4 0 0 0 0 ... 0.0983 \n", |
|
|
3674 |
".. ... ... ... ... ... ... \n", |
|
|
3675 |
"368 2 2 2 2 ... -0.0291 \n", |
|
|
3676 |
"369 0 0 0 0 ... 0.0497 \n", |
|
|
3677 |
"370 1 1 1 1 ... 0.3822 \n", |
|
|
3678 |
"371 0 0 0 0 ... -0.6853 \n", |
|
|
3679 |
"372 0 0 0 0 ... 0.0517 \n", |
|
|
3680 |
"\n", |
|
|
3681 |
" ZXDA_rnaseq ZXDB_rnaseq ZXDC_rnaseq ZYG11A_rnaseq ZYG11B_rnaseq \\\n", |
|
|
3682 |
"0 4.1375 3.9664 1.8437 -0.3959 -0.2561 \n", |
|
|
3683 |
"1 0.3393 0.2769 1.7320 -0.0975 2.6955 \n", |
|
|
3684 |
"2 1.6846 0.7711 -0.3061 -0.5016 2.8548 \n", |
|
|
3685 |
"3 0.8344 1.5075 3.6068 -0.5004 -0.0747 \n", |
|
|
3686 |
"4 -0.7908 -0.0053 -0.0643 -0.3706 0.3870 \n", |
|
|
3687 |
".. ... ... ... ... ... \n", |
|
|
3688 |
"368 -0.1058 -0.6721 0.2802 1.9504 -0.8784 \n", |
|
|
3689 |
"369 0.3673 -0.2208 0.3034 3.2580 -0.2089 \n", |
|
|
3690 |
"370 -0.7003 -0.7661 -1.7035 -0.5423 -0.3488 \n", |
|
|
3691 |
"371 -1.0240 -1.2890 -1.5666 -0.1270 -1.4662 \n", |
|
|
3692 |
"372 -0.3570 -0.4843 -0.3792 -0.1964 0.4200 \n", |
|
|
3693 |
"\n", |
|
|
3694 |
" ZYX_rnaseq ZZEF1_rnaseq ZZZ3_rnaseq TPTEP1_rnaseq \n", |
|
|
3695 |
"0 -0.2866 1.8770 -0.3179 -0.3633 \n", |
|
|
3696 |
"1 -0.6741 1.0323 1.2766 -0.3982 \n", |
|
|
3697 |
"2 -0.6171 -0.8608 -0.0486 -0.3962 \n", |
|
|
3698 |
"3 -0.2185 -0.4379 1.6913 1.7748 \n", |
|
|
3699 |
"4 -0.5589 -0.5979 0.0047 -0.3548 \n", |
|
|
3700 |
".. ... ... ... ... \n", |
|
|
3701 |
"368 0.9506 0.0607 1.1883 -0.3521 \n", |
|
|
3702 |
"369 1.6053 -0.8746 -0.4491 -0.3450 \n", |
|
|
3703 |
"370 1.3713 -0.4365 2.3456 -0.3866 \n", |
|
|
3704 |
"371 0.3981 -0.5976 -1.3822 -0.4157 \n", |
|
|
3705 |
"372 3.2547 -0.1232 3.4519 -0.1962 \n", |
|
|
3706 |
"\n", |
|
|
3707 |
"[373 rows x 20395 columns]" |
|
|
3708 |
] |
|
|
3709 |
}, |
|
|
3710 |
"execution_count": 68, |
|
|
3711 |
"metadata": {}, |
|
|
3712 |
"output_type": "execute_result" |
|
|
3713 |
} |
|
|
3714 |
], |
|
|
3715 |
"source": [ |
|
|
3716 |
"genomic_features" |
|
|
3717 |
] |
|
|
3718 |
}, |
|
|
3719 |
{ |
|
|
3720 |
"cell_type": "code", |
|
|
3721 |
"execution_count": 11, |
|
|
3722 |
"metadata": {}, |
|
|
3723 |
"outputs": [], |
|
|
3724 |
"source": [ |
|
|
3725 |
"if 'case_id' not in slide_data:\n", |
|
|
3726 |
" slide_data.index = slide_data.index.str[:12]\n", |
|
|
3727 |
" slide_data['case_id'] = slide_data.index\n", |
|
|
3728 |
" slide_data = slide_data.reset_index(drop=True)" |
|
|
3729 |
] |
|
|
3730 |
}, |
|
|
3731 |
{ |
|
|
3732 |
"cell_type": "code", |
|
|
3733 |
"execution_count": 14, |
|
|
3734 |
"metadata": {}, |
|
|
3735 |
"outputs": [], |
|
|
3736 |
"source": [ |
|
|
3737 |
"new_cols = list(slide_data.columns[-2:]) + list(slide_data.columns[:-2])\n", |
|
|
3738 |
"slide_data = slide_data[new_cols]" |
|
|
3739 |
] |
|
|
3740 |
}, |
|
|
3741 |
{ |
|
|
3742 |
"cell_type": "code", |
|
|
3743 |
"execution_count": 15, |
|
|
3744 |
"metadata": {}, |
|
|
3745 |
"outputs": [ |
|
|
3746 |
{ |
|
|
3747 |
"data": { |
|
|
3748 |
"text/html": [ |
|
|
3749 |
"<div>\n", |
|
|
3750 |
"<style scoped>\n", |
|
|
3751 |
" .dataframe tbody tr th:only-of-type {\n", |
|
|
3752 |
" vertical-align: middle;\n", |
|
|
3753 |
" }\n", |
|
|
3754 |
"\n", |
|
|
3755 |
" .dataframe tbody tr th {\n", |
|
|
3756 |
" vertical-align: top;\n", |
|
|
3757 |
" }\n", |
|
|
3758 |
"\n", |
|
|
3759 |
" .dataframe thead th {\n", |
|
|
3760 |
" text-align: right;\n", |
|
|
3761 |
" }\n", |
|
|
3762 |
"</style>\n", |
|
|
3763 |
"<table border=\"1\" class=\"dataframe\">\n", |
|
|
3764 |
" <thead>\n", |
|
|
3765 |
" <tr style=\"text-align: right;\">\n", |
|
|
3766 |
" <th></th>\n", |
|
|
3767 |
" <th>ZZZ3_rnaseq</th>\n", |
|
|
3768 |
" <th>TPTEP1_rnaseq</th>\n", |
|
|
3769 |
" <th>slide_id</th>\n", |
|
|
3770 |
" <th>site</th>\n", |
|
|
3771 |
" <th>is_female</th>\n", |
|
|
3772 |
" <th>oncotree_code</th>\n", |
|
|
3773 |
" <th>age</th>\n", |
|
|
3774 |
" <th>survival_months</th>\n", |
|
|
3775 |
" <th>censorship</th>\n", |
|
|
3776 |
" <th>train</th>\n", |
|
|
3777 |
" <th>...</th>\n", |
|
|
3778 |
" <th>ZW10_rnaseq</th>\n", |
|
|
3779 |
" <th>ZWILCH_rnaseq</th>\n", |
|
|
3780 |
" <th>ZWINT_rnaseq</th>\n", |
|
|
3781 |
" <th>ZXDA_rnaseq</th>\n", |
|
|
3782 |
" <th>ZXDB_rnaseq</th>\n", |
|
|
3783 |
" <th>ZXDC_rnaseq</th>\n", |
|
|
3784 |
" <th>ZYG11A_rnaseq</th>\n", |
|
|
3785 |
" <th>ZYG11B_rnaseq</th>\n", |
|
|
3786 |
" <th>ZYX_rnaseq</th>\n", |
|
|
3787 |
" <th>ZZEF1_rnaseq</th>\n", |
|
|
3788 |
" </tr>\n", |
|
|
3789 |
" <tr>\n", |
|
|
3790 |
" <th>case_id</th>\n", |
|
|
3791 |
" <th></th>\n", |
|
|
3792 |
" <th></th>\n", |
|
|
3793 |
" <th></th>\n", |
|
|
3794 |
" <th></th>\n", |
|
|
3795 |
" <th></th>\n", |
|
|
3796 |
" <th></th>\n", |
|
|
3797 |
" <th></th>\n", |
|
|
3798 |
" <th></th>\n", |
|
|
3799 |
" <th></th>\n", |
|
|
3800 |
" <th></th>\n", |
|
|
3801 |
" <th></th>\n", |
|
|
3802 |
" <th></th>\n", |
|
|
3803 |
" <th></th>\n", |
|
|
3804 |
" <th></th>\n", |
|
|
3805 |
" <th></th>\n", |
|
|
3806 |
" <th></th>\n", |
|
|
3807 |
" <th></th>\n", |
|
|
3808 |
" <th></th>\n", |
|
|
3809 |
" <th></th>\n", |
|
|
3810 |
" <th></th>\n", |
|
|
3811 |
" <th></th>\n", |
|
|
3812 |
" </tr>\n", |
|
|
3813 |
" </thead>\n", |
|
|
3814 |
" <tbody>\n", |
|
|
3815 |
" <tr>\n", |
|
|
3816 |
" <th>TCGA-2F-A9KO</th>\n", |
|
|
3817 |
" <td>-0.3179</td>\n", |
|
|
3818 |
" <td>-0.3633</td>\n", |
|
|
3819 |
" <td>TCGA-2F-A9KO-01Z-00-DX1.195576CF-B739-4BD9-B15...</td>\n", |
|
|
3820 |
" <td>2F</td>\n", |
|
|
3821 |
" <td>0</td>\n", |
|
|
3822 |
" <td>BLCA</td>\n", |
|
|
3823 |
" <td>63</td>\n", |
|
|
3824 |
" <td>24.11</td>\n", |
|
|
3825 |
" <td>0</td>\n", |
|
|
3826 |
" <td>1.0</td>\n", |
|
|
3827 |
" <td>...</td>\n", |
|
|
3828 |
" <td>-0.7172</td>\n", |
|
|
3829 |
" <td>0.7409</td>\n", |
|
|
3830 |
" <td>-0.8388</td>\n", |
|
|
3831 |
" <td>4.1375</td>\n", |
|
|
3832 |
" <td>3.9664</td>\n", |
|
|
3833 |
" <td>1.8437</td>\n", |
|
|
3834 |
" <td>-0.3959</td>\n", |
|
|
3835 |
" <td>-0.2561</td>\n", |
|
|
3836 |
" <td>-0.2866</td>\n", |
|
|
3837 |
" <td>1.8770</td>\n", |
|
|
3838 |
" </tr>\n", |
|
|
3839 |
" <tr>\n", |
|
|
3840 |
" <th>TCGA-2F-A9KP</th>\n", |
|
|
3841 |
" <td>1.2766</td>\n", |
|
|
3842 |
" <td>-0.3982</td>\n", |
|
|
3843 |
" <td>TCGA-2F-A9KP-01Z-00-DX1.3CDF534E-958F-4467-AA7...</td>\n", |
|
|
3844 |
" <td>2F</td>\n", |
|
|
3845 |
" <td>0</td>\n", |
|
|
3846 |
" <td>BLCA</td>\n", |
|
|
3847 |
" <td>66</td>\n", |
|
|
3848 |
" <td>11.96</td>\n", |
|
|
3849 |
" <td>0</td>\n", |
|
|
3850 |
" <td>1.0</td>\n", |
|
|
3851 |
" <td>...</td>\n", |
|
|
3852 |
" <td>0.6373</td>\n", |
|
|
3853 |
" <td>0.8559</td>\n", |
|
|
3854 |
" <td>-0.1083</td>\n", |
|
|
3855 |
" <td>0.3393</td>\n", |
|
|
3856 |
" <td>0.2769</td>\n", |
|
|
3857 |
" <td>1.7320</td>\n", |
|
|
3858 |
" <td>-0.0975</td>\n", |
|
|
3859 |
" <td>2.6955</td>\n", |
|
|
3860 |
" <td>-0.6741</td>\n", |
|
|
3861 |
" <td>1.0323</td>\n", |
|
|
3862 |
" </tr>\n", |
|
|
3863 |
" <tr>\n", |
|
|
3864 |
" <th>TCGA-2F-A9KP</th>\n", |
|
|
3865 |
" <td>1.2766</td>\n", |
|
|
3866 |
" <td>-0.3982</td>\n", |
|
|
3867 |
" <td>TCGA-2F-A9KP-01Z-00-DX2.718C82A3-252B-498E-BFB...</td>\n", |
|
|
3868 |
" <td>2F</td>\n", |
|
|
3869 |
" <td>0</td>\n", |
|
|
3870 |
" <td>BLCA</td>\n", |
|
|
3871 |
" <td>66</td>\n", |
|
|
3872 |
" <td>11.96</td>\n", |
|
|
3873 |
" <td>0</td>\n", |
|
|
3874 |
" <td>1.0</td>\n", |
|
|
3875 |
" <td>...</td>\n", |
|
|
3876 |
" <td>0.6373</td>\n", |
|
|
3877 |
" <td>0.8559</td>\n", |
|
|
3878 |
" <td>-0.1083</td>\n", |
|
|
3879 |
" <td>0.3393</td>\n", |
|
|
3880 |
" <td>0.2769</td>\n", |
|
|
3881 |
" <td>1.7320</td>\n", |
|
|
3882 |
" <td>-0.0975</td>\n", |
|
|
3883 |
" <td>2.6955</td>\n", |
|
|
3884 |
" <td>-0.6741</td>\n", |
|
|
3885 |
" <td>1.0323</td>\n", |
|
|
3886 |
" </tr>\n", |
|
|
3887 |
" <tr>\n", |
|
|
3888 |
" <th>TCGA-2F-A9KQ</th>\n", |
|
|
3889 |
" <td>-0.0486</td>\n", |
|
|
3890 |
" <td>-0.3962</td>\n", |
|
|
3891 |
" <td>TCGA-2F-A9KQ-01Z-00-DX1.1C8CB2DD-5CC6-4E99-A0F...</td>\n", |
|
|
3892 |
" <td>2F</td>\n", |
|
|
3893 |
" <td>0</td>\n", |
|
|
3894 |
" <td>BLCA</td>\n", |
|
|
3895 |
" <td>69</td>\n", |
|
|
3896 |
" <td>94.81</td>\n", |
|
|
3897 |
" <td>1</td>\n", |
|
|
3898 |
" <td>1.0</td>\n", |
|
|
3899 |
" <td>...</td>\n", |
|
|
3900 |
" <td>-0.5676</td>\n", |
|
|
3901 |
" <td>-0.0621</td>\n", |
|
|
3902 |
" <td>-0.4155</td>\n", |
|
|
3903 |
" <td>1.6846</td>\n", |
|
|
3904 |
" <td>0.7711</td>\n", |
|
|
3905 |
" <td>-0.3061</td>\n", |
|
|
3906 |
" <td>-0.5016</td>\n", |
|
|
3907 |
" <td>2.8548</td>\n", |
|
|
3908 |
" <td>-0.6171</td>\n", |
|
|
3909 |
" <td>-0.8608</td>\n", |
|
|
3910 |
" </tr>\n", |
|
|
3911 |
" <tr>\n", |
|
|
3912 |
" <th>TCGA-2F-A9KR</th>\n", |
|
|
3913 |
" <td>1.6913</td>\n", |
|
|
3914 |
" <td>1.7748</td>\n", |
|
|
3915 |
" <td>TCGA-2F-A9KR-01Z-00-DX1.D6A4BD2D-18F3-4FA6-827...</td>\n", |
|
|
3916 |
" <td>2F</td>\n", |
|
|
3917 |
" <td>1</td>\n", |
|
|
3918 |
" <td>BLCA</td>\n", |
|
|
3919 |
" <td>59</td>\n", |
|
|
3920 |
" <td>104.57</td>\n", |
|
|
3921 |
" <td>0</td>\n", |
|
|
3922 |
" <td>1.0</td>\n", |
|
|
3923 |
" <td>...</td>\n", |
|
|
3924 |
" <td>-1.3825</td>\n", |
|
|
3925 |
" <td>0.3550</td>\n", |
|
|
3926 |
" <td>-0.8143</td>\n", |
|
|
3927 |
" <td>0.8344</td>\n", |
|
|
3928 |
" <td>1.5075</td>\n", |
|
|
3929 |
" <td>3.6068</td>\n", |
|
|
3930 |
" <td>-0.5004</td>\n", |
|
|
3931 |
" <td>-0.0747</td>\n", |
|
|
3932 |
" <td>-0.2185</td>\n", |
|
|
3933 |
" <td>-0.4379</td>\n", |
|
|
3934 |
" </tr>\n", |
|
|
3935 |
" <tr>\n", |
|
|
3936 |
" <th>...</th>\n", |
|
|
3937 |
" <td>...</td>\n", |
|
|
3938 |
" <td>...</td>\n", |
|
|
3939 |
" <td>...</td>\n", |
|
|
3940 |
" <td>...</td>\n", |
|
|
3941 |
" <td>...</td>\n", |
|
|
3942 |
" <td>...</td>\n", |
|
|
3943 |
" <td>...</td>\n", |
|
|
3944 |
" <td>...</td>\n", |
|
|
3945 |
" <td>...</td>\n", |
|
|
3946 |
" <td>...</td>\n", |
|
|
3947 |
" <td>...</td>\n", |
|
|
3948 |
" <td>...</td>\n", |
|
|
3949 |
" <td>...</td>\n", |
|
|
3950 |
" <td>...</td>\n", |
|
|
3951 |
" <td>...</td>\n", |
|
|
3952 |
" <td>...</td>\n", |
|
|
3953 |
" <td>...</td>\n", |
|
|
3954 |
" <td>...</td>\n", |
|
|
3955 |
" <td>...</td>\n", |
|
|
3956 |
" <td>...</td>\n", |
|
|
3957 |
" <td>...</td>\n", |
|
|
3958 |
" </tr>\n", |
|
|
3959 |
" <tr>\n", |
|
|
3960 |
" <th>TCGA-ZF-AA54</th>\n", |
|
|
3961 |
" <td>1.1883</td>\n", |
|
|
3962 |
" <td>-0.3521</td>\n", |
|
|
3963 |
" <td>TCGA-ZF-AA54-01Z-00-DX1.9118BB51-333A-4257-A79...</td>\n", |
|
|
3964 |
" <td>ZF</td>\n", |
|
|
3965 |
" <td>0</td>\n", |
|
|
3966 |
" <td>BLCA</td>\n", |
|
|
3967 |
" <td>71</td>\n", |
|
|
3968 |
" <td>19.38</td>\n", |
|
|
3969 |
" <td>0</td>\n", |
|
|
3970 |
" <td>1.0</td>\n", |
|
|
3971 |
" <td>...</td>\n", |
|
|
3972 |
" <td>-0.0898</td>\n", |
|
|
3973 |
" <td>2.1092</td>\n", |
|
|
3974 |
" <td>-0.0291</td>\n", |
|
|
3975 |
" <td>-0.1058</td>\n", |
|
|
3976 |
" <td>-0.6721</td>\n", |
|
|
3977 |
" <td>0.2802</td>\n", |
|
|
3978 |
" <td>1.9504</td>\n", |
|
|
3979 |
" <td>-0.8784</td>\n", |
|
|
3980 |
" <td>0.9506</td>\n", |
|
|
3981 |
" <td>0.0607</td>\n", |
|
|
3982 |
" </tr>\n", |
|
|
3983 |
" <tr>\n", |
|
|
3984 |
" <th>TCGA-ZF-AA58</th>\n", |
|
|
3985 |
" <td>-0.4491</td>\n", |
|
|
3986 |
" <td>-0.3450</td>\n", |
|
|
3987 |
" <td>TCGA-ZF-AA58-01Z-00-DX1.85C3611E-11FA-4AAE-B88...</td>\n", |
|
|
3988 |
" <td>ZF</td>\n", |
|
|
3989 |
" <td>1</td>\n", |
|
|
3990 |
" <td>BLCA</td>\n", |
|
|
3991 |
" <td>61</td>\n", |
|
|
3992 |
" <td>54.17</td>\n", |
|
|
3993 |
" <td>1</td>\n", |
|
|
3994 |
" <td>1.0</td>\n", |
|
|
3995 |
" <td>...</td>\n", |
|
|
3996 |
" <td>-0.2075</td>\n", |
|
|
3997 |
" <td>-0.0617</td>\n", |
|
|
3998 |
" <td>0.0497</td>\n", |
|
|
3999 |
" <td>0.3673</td>\n", |
|
|
4000 |
" <td>-0.2208</td>\n", |
|
|
4001 |
" <td>0.3034</td>\n", |
|
|
4002 |
" <td>3.2580</td>\n", |
|
|
4003 |
" <td>-0.2089</td>\n", |
|
|
4004 |
" <td>1.6053</td>\n", |
|
|
4005 |
" <td>-0.8746</td>\n", |
|
|
4006 |
" </tr>\n", |
|
|
4007 |
" <tr>\n", |
|
|
4008 |
" <th>TCGA-ZF-AA5H</th>\n", |
|
|
4009 |
" <td>2.3456</td>\n", |
|
|
4010 |
" <td>-0.3866</td>\n", |
|
|
4011 |
" <td>TCGA-ZF-AA5H-01Z-00-DX1.2B5DF00E-E0FD-4C58-A82...</td>\n", |
|
|
4012 |
" <td>ZF</td>\n", |
|
|
4013 |
" <td>1</td>\n", |
|
|
4014 |
" <td>BLCA</td>\n", |
|
|
4015 |
" <td>60</td>\n", |
|
|
4016 |
" <td>29.47</td>\n", |
|
|
4017 |
" <td>1</td>\n", |
|
|
4018 |
" <td>1.0</td>\n", |
|
|
4019 |
" <td>...</td>\n", |
|
|
4020 |
" <td>-1.4118</td>\n", |
|
|
4021 |
" <td>-0.1236</td>\n", |
|
|
4022 |
" <td>0.3822</td>\n", |
|
|
4023 |
" <td>-0.7003</td>\n", |
|
|
4024 |
" <td>-0.7661</td>\n", |
|
|
4025 |
" <td>-1.7035</td>\n", |
|
|
4026 |
" <td>-0.5423</td>\n", |
|
|
4027 |
" <td>-0.3488</td>\n", |
|
|
4028 |
" <td>1.3713</td>\n", |
|
|
4029 |
" <td>-0.4365</td>\n", |
|
|
4030 |
" </tr>\n", |
|
|
4031 |
" <tr>\n", |
|
|
4032 |
" <th>TCGA-ZF-AA5N</th>\n", |
|
|
4033 |
" <td>-1.3822</td>\n", |
|
|
4034 |
" <td>-0.4157</td>\n", |
|
|
4035 |
" <td>TCGA-ZF-AA5N-01Z-00-DX1.A207E3EE-CC7D-4267-A77...</td>\n", |
|
|
4036 |
" <td>ZF</td>\n", |
|
|
4037 |
" <td>1</td>\n", |
|
|
4038 |
" <td>BLCA</td>\n", |
|
|
4039 |
" <td>62</td>\n", |
|
|
4040 |
" <td>5.52</td>\n", |
|
|
4041 |
" <td>0</td>\n", |
|
|
4042 |
" <td>1.0</td>\n", |
|
|
4043 |
" <td>...</td>\n", |
|
|
4044 |
" <td>-0.1733</td>\n", |
|
|
4045 |
" <td>-0.2397</td>\n", |
|
|
4046 |
" <td>-0.6853</td>\n", |
|
|
4047 |
" <td>-1.0240</td>\n", |
|
|
4048 |
" <td>-1.2890</td>\n", |
|
|
4049 |
" <td>-1.5666</td>\n", |
|
|
4050 |
" <td>-0.1270</td>\n", |
|
|
4051 |
" <td>-1.4662</td>\n", |
|
|
4052 |
" <td>0.3981</td>\n", |
|
|
4053 |
" <td>-0.5976</td>\n", |
|
|
4054 |
" </tr>\n", |
|
|
4055 |
" <tr>\n", |
|
|
4056 |
" <th>TCGA-ZF-AA5P</th>\n", |
|
|
4057 |
" <td>3.4519</td>\n", |
|
|
4058 |
" <td>-0.1962</td>\n", |
|
|
4059 |
" <td>TCGA-ZF-AA5P-01Z-00-DX1.B91697A2-A186-4E67-A81...</td>\n", |
|
|
4060 |
" <td>ZF</td>\n", |
|
|
4061 |
" <td>0</td>\n", |
|
|
4062 |
" <td>BLCA</td>\n", |
|
|
4063 |
" <td>65</td>\n", |
|
|
4064 |
" <td>12.22</td>\n", |
|
|
4065 |
" <td>1</td>\n", |
|
|
4066 |
" <td>1.0</td>\n", |
|
|
4067 |
" <td>...</td>\n", |
|
|
4068 |
" <td>-1.1056</td>\n", |
|
|
4069 |
" <td>-0.6634</td>\n", |
|
|
4070 |
" <td>0.0517</td>\n", |
|
|
4071 |
" <td>-0.3570</td>\n", |
|
|
4072 |
" <td>-0.4843</td>\n", |
|
|
4073 |
" <td>-0.3792</td>\n", |
|
|
4074 |
" <td>-0.1964</td>\n", |
|
|
4075 |
" <td>0.4200</td>\n", |
|
|
4076 |
" <td>3.2547</td>\n", |
|
|
4077 |
" <td>-0.1232</td>\n", |
|
|
4078 |
" </tr>\n", |
|
|
4079 |
" </tbody>\n", |
|
|
4080 |
"</table>\n", |
|
|
4081 |
"<p>437 rows × 20403 columns</p>\n", |
|
|
4082 |
"</div>" |
|
|
4083 |
], |
|
|
4084 |
"text/plain": [ |
|
|
4085 |
" ZZZ3_rnaseq TPTEP1_rnaseq \\\n", |
|
|
4086 |
"case_id \n", |
|
|
4087 |
"TCGA-2F-A9KO -0.3179 -0.3633 \n", |
|
|
4088 |
"TCGA-2F-A9KP 1.2766 -0.3982 \n", |
|
|
4089 |
"TCGA-2F-A9KP 1.2766 -0.3982 \n", |
|
|
4090 |
"TCGA-2F-A9KQ -0.0486 -0.3962 \n", |
|
|
4091 |
"TCGA-2F-A9KR 1.6913 1.7748 \n", |
|
|
4092 |
"... ... ... \n", |
|
|
4093 |
"TCGA-ZF-AA54 1.1883 -0.3521 \n", |
|
|
4094 |
"TCGA-ZF-AA58 -0.4491 -0.3450 \n", |
|
|
4095 |
"TCGA-ZF-AA5H 2.3456 -0.3866 \n", |
|
|
4096 |
"TCGA-ZF-AA5N -1.3822 -0.4157 \n", |
|
|
4097 |
"TCGA-ZF-AA5P 3.4519 -0.1962 \n", |
|
|
4098 |
"\n", |
|
|
4099 |
" slide_id site \\\n", |
|
|
4100 |
"case_id \n", |
|
|
4101 |
"TCGA-2F-A9KO TCGA-2F-A9KO-01Z-00-DX1.195576CF-B739-4BD9-B15... 2F \n", |
|
|
4102 |
"TCGA-2F-A9KP TCGA-2F-A9KP-01Z-00-DX1.3CDF534E-958F-4467-AA7... 2F \n", |
|
|
4103 |
"TCGA-2F-A9KP TCGA-2F-A9KP-01Z-00-DX2.718C82A3-252B-498E-BFB... 2F \n", |
|
|
4104 |
"TCGA-2F-A9KQ TCGA-2F-A9KQ-01Z-00-DX1.1C8CB2DD-5CC6-4E99-A0F... 2F \n", |
|
|
4105 |
"TCGA-2F-A9KR TCGA-2F-A9KR-01Z-00-DX1.D6A4BD2D-18F3-4FA6-827... 2F \n", |
|
|
4106 |
"... ... ... \n", |
|
|
4107 |
"TCGA-ZF-AA54 TCGA-ZF-AA54-01Z-00-DX1.9118BB51-333A-4257-A79... ZF \n", |
|
|
4108 |
"TCGA-ZF-AA58 TCGA-ZF-AA58-01Z-00-DX1.85C3611E-11FA-4AAE-B88... ZF \n", |
|
|
4109 |
"TCGA-ZF-AA5H TCGA-ZF-AA5H-01Z-00-DX1.2B5DF00E-E0FD-4C58-A82... ZF \n", |
|
|
4110 |
"TCGA-ZF-AA5N TCGA-ZF-AA5N-01Z-00-DX1.A207E3EE-CC7D-4267-A77... ZF \n", |
|
|
4111 |
"TCGA-ZF-AA5P TCGA-ZF-AA5P-01Z-00-DX1.B91697A2-A186-4E67-A81... ZF \n", |
|
|
4112 |
"\n", |
|
|
4113 |
" is_female oncotree_code age survival_months censorship \\\n", |
|
|
4114 |
"case_id \n", |
|
|
4115 |
"TCGA-2F-A9KO 0 BLCA 63 24.11 0 \n", |
|
|
4116 |
"TCGA-2F-A9KP 0 BLCA 66 11.96 0 \n", |
|
|
4117 |
"TCGA-2F-A9KP 0 BLCA 66 11.96 0 \n", |
|
|
4118 |
"TCGA-2F-A9KQ 0 BLCA 69 94.81 1 \n", |
|
|
4119 |
"TCGA-2F-A9KR 1 BLCA 59 104.57 0 \n", |
|
|
4120 |
"... ... ... ... ... ... \n", |
|
|
4121 |
"TCGA-ZF-AA54 0 BLCA 71 19.38 0 \n", |
|
|
4122 |
"TCGA-ZF-AA58 1 BLCA 61 54.17 1 \n", |
|
|
4123 |
"TCGA-ZF-AA5H 1 BLCA 60 29.47 1 \n", |
|
|
4124 |
"TCGA-ZF-AA5N 1 BLCA 62 5.52 0 \n", |
|
|
4125 |
"TCGA-ZF-AA5P 0 BLCA 65 12.22 1 \n", |
|
|
4126 |
"\n", |
|
|
4127 |
" train ... ZW10_rnaseq ZWILCH_rnaseq ZWINT_rnaseq \\\n", |
|
|
4128 |
"case_id ... \n", |
|
|
4129 |
"TCGA-2F-A9KO 1.0 ... -0.7172 0.7409 -0.8388 \n", |
|
|
4130 |
"TCGA-2F-A9KP 1.0 ... 0.6373 0.8559 -0.1083 \n", |
|
|
4131 |
"TCGA-2F-A9KP 1.0 ... 0.6373 0.8559 -0.1083 \n", |
|
|
4132 |
"TCGA-2F-A9KQ 1.0 ... -0.5676 -0.0621 -0.4155 \n", |
|
|
4133 |
"TCGA-2F-A9KR 1.0 ... -1.3825 0.3550 -0.8143 \n", |
|
|
4134 |
"... ... ... ... ... ... \n", |
|
|
4135 |
"TCGA-ZF-AA54 1.0 ... -0.0898 2.1092 -0.0291 \n", |
|
|
4136 |
"TCGA-ZF-AA58 1.0 ... -0.2075 -0.0617 0.0497 \n", |
|
|
4137 |
"TCGA-ZF-AA5H 1.0 ... -1.4118 -0.1236 0.3822 \n", |
|
|
4138 |
"TCGA-ZF-AA5N 1.0 ... -0.1733 -0.2397 -0.6853 \n", |
|
|
4139 |
"TCGA-ZF-AA5P 1.0 ... -1.1056 -0.6634 0.0517 \n", |
|
|
4140 |
"\n", |
|
|
4141 |
" ZXDA_rnaseq ZXDB_rnaseq ZXDC_rnaseq ZYG11A_rnaseq \\\n", |
|
|
4142 |
"case_id \n", |
|
|
4143 |
"TCGA-2F-A9KO 4.1375 3.9664 1.8437 -0.3959 \n", |
|
|
4144 |
"TCGA-2F-A9KP 0.3393 0.2769 1.7320 -0.0975 \n", |
|
|
4145 |
"TCGA-2F-A9KP 0.3393 0.2769 1.7320 -0.0975 \n", |
|
|
4146 |
"TCGA-2F-A9KQ 1.6846 0.7711 -0.3061 -0.5016 \n", |
|
|
4147 |
"TCGA-2F-A9KR 0.8344 1.5075 3.6068 -0.5004 \n", |
|
|
4148 |
"... ... ... ... ... \n", |
|
|
4149 |
"TCGA-ZF-AA54 -0.1058 -0.6721 0.2802 1.9504 \n", |
|
|
4150 |
"TCGA-ZF-AA58 0.3673 -0.2208 0.3034 3.2580 \n", |
|
|
4151 |
"TCGA-ZF-AA5H -0.7003 -0.7661 -1.7035 -0.5423 \n", |
|
|
4152 |
"TCGA-ZF-AA5N -1.0240 -1.2890 -1.5666 -0.1270 \n", |
|
|
4153 |
"TCGA-ZF-AA5P -0.3570 -0.4843 -0.3792 -0.1964 \n", |
|
|
4154 |
"\n", |
|
|
4155 |
" ZYG11B_rnaseq ZYX_rnaseq ZZEF1_rnaseq \n", |
|
|
4156 |
"case_id \n", |
|
|
4157 |
"TCGA-2F-A9KO -0.2561 -0.2866 1.8770 \n", |
|
|
4158 |
"TCGA-2F-A9KP 2.6955 -0.6741 1.0323 \n", |
|
|
4159 |
"TCGA-2F-A9KP 2.6955 -0.6741 1.0323 \n", |
|
|
4160 |
"TCGA-2F-A9KQ 2.8548 -0.6171 -0.8608 \n", |
|
|
4161 |
"TCGA-2F-A9KR -0.0747 -0.2185 -0.4379 \n", |
|
|
4162 |
"... ... ... ... \n", |
|
|
4163 |
"TCGA-ZF-AA54 -0.8784 0.9506 0.0607 \n", |
|
|
4164 |
"TCGA-ZF-AA58 -0.2089 1.6053 -0.8746 \n", |
|
|
4165 |
"TCGA-ZF-AA5H -0.3488 1.3713 -0.4365 \n", |
|
|
4166 |
"TCGA-ZF-AA5N -1.4662 0.3981 -0.5976 \n", |
|
|
4167 |
"TCGA-ZF-AA5P 0.4200 3.2547 -0.1232 \n", |
|
|
4168 |
"\n", |
|
|
4169 |
"[437 rows x 20403 columns]" |
|
|
4170 |
] |
|
|
4171 |
}, |
|
|
4172 |
"execution_count": 15, |
|
|
4173 |
"metadata": {}, |
|
|
4174 |
"output_type": "execute_result" |
|
|
4175 |
} |
|
|
4176 |
], |
|
|
4177 |
"source": [ |
|
|
4178 |
"slide_data" |
|
|
4179 |
] |
|
|
4180 |
} |
|
|
4181 |
], |
|
|
4182 |
"metadata": { |
|
|
4183 |
"kernelspec": { |
|
|
4184 |
"display_name": "Python 3", |
|
|
4185 |
"language": "python", |
|
|
4186 |
"name": "python3" |
|
|
4187 |
}, |
|
|
4188 |
"language_info": { |
|
|
4189 |
"codemirror_mode": { |
|
|
4190 |
"name": "ipython", |
|
|
4191 |
"version": 3 |
|
|
4192 |
}, |
|
|
4193 |
"file_extension": ".py", |
|
|
4194 |
"mimetype": "text/x-python", |
|
|
4195 |
"name": "python", |
|
|
4196 |
"nbconvert_exporter": "python", |
|
|
4197 |
"pygments_lexer": "ipython3", |
|
|
4198 |
"version": "3.8.3" |
|
|
4199 |
} |
|
|
4200 |
}, |
|
|
4201 |
"nbformat": 4, |
|
|
4202 |
"nbformat_minor": 4 |
|
|
4203 |
} |