Datasets
Models
Downloads: 1
[7b3b0e]:
/
linear_cox.ipynb
335 lines (334 with data), 21.3 kB
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Packages required\n",
"- numpy==1.14.0\n",
"- matplotlib==3.0.3\n",
"- pandas==0.24.2\n",
"- nibabel==2.5.0\n",
"- scikit-learn\n",
"- scikit-survival 0.12 with ``pip install scikit-survival==0.12``\n",
"- lifelines 0.23.9 with ``pip install lifelines``"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import sksurv\n",
"import numpy as np\n",
"%matplotlib inline\n",
"from matplotlib import pyplot as plt\n",
"\n",
"from sksurv.linear_model import CoxPHSurvivalAnalysis\n",
"from lifelines.utils import concordance_index\n",
"\n",
"from utils import DataLoader, get_structured_array, run_coxnet"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading data for mode valid from location data/stanford/labels/recurrence_labels/\n",
"Loading data for mode test from location data/stanford/labels/recurrence_labels/\n",
"Fold : 0 || Training set: (74,) | Validation set: (11,) | Test set: (22,)\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading data for mode valid from location data/stanford/labels/recurrence_labels/\n",
"Loading data for mode test from location data/stanford/labels/recurrence_labels/\n",
"Fold : 1 || Training set: (74,) | Validation set: (11,) | Test set: (22,)\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading data for mode valid from location data/stanford/labels/recurrence_labels/\n",
"Loading data for mode test from location data/stanford/labels/recurrence_labels/\n",
"Fold : 2 || Training set: (75,) | Validation set: (11,) | Test set: (21,)\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading data for mode valid from location data/stanford/labels/recurrence_labels/\n",
"Loading data for mode test from location data/stanford/labels/recurrence_labels/\n",
"Fold : 3 || Training set: (75,) | Validation set: (11,) | Test set: (21,)\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading data for mode valid from location data/stanford/labels/recurrence_labels/\n",
"Loading data for mode test from location data/stanford/labels/recurrence_labels/\n",
"Fold : 4 || Training set: (75,) | Validation set: (11,) | Test set: (21,)\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n",
"/home/vs5/.virtualenvs/radiogenomics/lib/python3.5/site-packages/sksurv/linear_model/coxnet.py:211: FutureWarning: The default value of alpha_min_ratio will depend on the sample size relative to the number of features in 0.13. If n_samples > n_features, the current default value 0.0001 will be used. If n_samples < n_features, 0.01 will be used instead.\n",
" FutureWarning)\n"
]
}
],
"source": [
"n_alphas = 100\n",
"l1_ratio = 0.3\n",
"num_genes = 500\n",
"\n",
"all_indices = []\n",
"\n",
"for fold_num in range(5):\n",
"\n",
" curr_indices = []\n",
"\n",
" data = DataLoader(fold=fold_num, num_genes=num_genes)\n",
" y_train = get_structured_array(data.y_train_bool, data.y_train_value)\n",
" y_valid = get_structured_array(data.y_valid_bool, data.y_valid_value)\n",
" y_test = get_structured_array(data.y_test_bool, data.y_test_value)\n",
" print('Fold : ' + str(fold_num) + ' || ' +\n",
" 'Training set: ' + str(y_train.shape) + ' | Validation set: ' + str(y_valid.shape) +\n",
" ' | Test set: ' + str(y_test.shape))\n",
"\n",
" ### Genomics\n",
"\n",
" gen_outputs, gen_scores = run_coxnet(l1_ratio, n_alphas,\n",
" data.gen_train , y_train,\n",
" data.gen_test , y_test)\n",
"\n",
" ### Pyradiomics\n",
"\n",
" pyrad_outputs, pyrad_scores = run_coxnet(l1_ratio, n_alphas,\n",
" data.pyrad_train , y_train,\n",
" data.pyrad_test , y_test)\n",
" \n",
" ### Densenet\n",
"\n",
" dense_outputs, dense_scores = run_coxnet(l1_ratio, n_alphas,\n",
" data.dense_train , y_train,\n",
" data.dense_test , y_test)\n",
"\n",
" ### Genomics-PyRadiomics\n",
"\n",
" feat1_train = np.concatenate((data.gen_train , \n",
" data.pyrad_train ), axis=1)\n",
" feat1_test = np.concatenate((data.gen_test , \n",
" data.pyrad_test ), axis=1)\n",
"\n",
" feat1_outputs, feat1_score = run_coxnet(l1_ratio, n_alphas,\n",
" feat1_train, y_train,\n",
" feat1_test, y_test)\n",
" feat1_hat_score = concordance_index(data.y_test_value ,\n",
" - gen_outputs - pyrad_outputs,\n",
" data.y_test_bool )\n",
" ### Genomics-DenseNet\n",
" \n",
" feat2_train = np.concatenate((data.gen_train , \n",
" data.dense_train ), axis=1)\n",
" feat2_test = np.concatenate((data.gen_test , \n",
" data.dense_test ), axis=1)\n",
"\n",
" feat2_outputs, feat2_score = run_coxnet(l1_ratio, n_alphas,\n",
" feat2_train, y_train,\n",
" feat2_test, y_test)\n",
" feat2_hat_score = concordance_index(data.y_test_value ,\n",
" - gen_outputs - dense_outputs,\n",
" data.y_test_bool )\n",
" \n",
" ### Genomics-PyRadiomics-DenseNet\n",
"\n",
" \n",
" feat3_train = np.concatenate((data.gen_train , \n",
" data.pyrad_train ,\n",
" data.dense_train ), axis=1)\n",
" feat3_test = np.concatenate((data.gen_test , \n",
" data.pyrad_test ,\n",
" data.dense_test ), axis=1)\n",
"\n",
" feat3_outputs, feat3_score = run_coxnet(l1_ratio, n_alphas,\n",
" feat3_train, y_train,\n",
" feat3_test, y_test)\n",
" feat3_hat_score = concordance_index(data.y_test_value ,\n",
" - gen_outputs - pyrad_outputs - dense_outputs,\n",
" data.y_test_bool )\n",
"\n",
"\n",
" curr_indices = [gen_scores, pyrad_scores, dense_scores,\n",
" feat1_score, feat1_hat_score,\n",
" feat2_score, feat2_hat_score,\n",
" feat3_score, feat3_hat_score]\n",
" all_indices.append(curr_indices)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"methods = ['genomics \\t ', 'pyradiomics \\t', 'densenet \\t',\n",
" 'gen-pyrad (ef) \\t', 'gen-pyrad (lf) \\t', 'gen-dense (ef) \\t', 'gen-dense (lf) \\t',\n",
" 'gen-pyrad-dense (ef)', 'gen-pyrad-dense (lf)',]"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"method \t\t \t|fd 1|fd 2|fd 3|fd 4|fd 5\n",
"-------------------------------------------------\n",
"genomics \t \t|0.52|0.45|0.60|0.72|0.75\n",
"pyradiomics \t\t|0.46|0.39|0.30|0.37|0.80\n",
"densenet \t\t|0.53|0.49|0.44|0.65|0.52\n",
"gen-pyrad (ef) \t\t|0.41|0.36|0.46|0.57|0.81\n",
"gen-pyrad (lf) \t\t|0.41|0.42|0.49|0.60|0.81\n",
"gen-dense (ef) \t\t|0.53|0.46|0.60|0.80|0.77\n",
"gen-dense (lf) \t\t|0.49|0.50|0.55|0.82|0.59\n",
"gen-pyrad-dense (ef)\t|0.43|0.37|0.47|0.61|0.82\n",
"gen-pyrad-dense (lf)\t|0.46|0.46|0.51|0.81|0.60\n"
]
}
],
"source": [
"print('method \\t\\t \\t|fd 1|fd 2|fd 3|fd 4|fd 5')\n",
"print('-------------------------------------------------')\n",
"for y in range(len(all_indices[0])):s\n",
" print(methods[y] + '\\t|' + '|'.join([\"{0:.2f}\".format(round(x[y], 2)) for x in all_indices]))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python3 (radiogenomics)",
"language": "python",
"name": "radiogenomics"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}