--- a
+++ b/AttentionMOI/src/preprocess.py
@@ -0,0 +1,78 @@
+import sys
+import pandas as pd
+
+
+def read_(file):
+    # read file
+    if file.endswith('.csv'):
+        df = pd.read_csv(file, index_col=0)
+    elif file.endswith('.csv.gz'):
+        df = pd.read_csv(file, compression='gzip', index_col=0)
+    else:
+        print('\n[Error]: The program cannot infer the format of {} . Currently, only the csv format is supported, please ensure that the file name suffix is .csv or .csv.gz.'.format(file))
+        sys.exit(0)
+    return df
+
+
+def read_omics(args):
+    omics = []
+    for file in args.omic_file:
+        df = read_(file)
+        df = df.fillna(0)  # fill nan with 0
+        omics.append(df)
+    return omics
+
+
+def read_label(args):
+    file = args.label_file
+    df = read_(file)
+    df = df.rename(
+        columns={df.columns.values[0]: 'label'})
+    return df
+
+
+def read_clin(args):
+    file = args.clin_file
+    df = None
+    if not file is None:
+        df = read_(file)
+        # fill na
+        df = df.fillna(0)
+    return df
+
+def process(df_omics, df_label, df_clin):
+    # extract patient id
+    patients = [df_tmp.index.to_list() for df_tmp in df_omics]
+    patients.append(df_label.index.to_list())
+    if not df_clin is None:
+        patients.append(df_clin.index.to_list())
+
+    # get shared patients between different data
+    patients_shared = patients[0]
+    for i in range(1, len(patients)):
+        patients_shared = list(set(patients_shared).intersection(patients[i]))
+
+    # extract shared patients' data
+    for i in range(len(df_omics)):
+        df_omics[i] = df_omics[i].loc[patients_shared, :].sort_index()
+    df_label = df_label.loc[patients_shared, :].sort_index()
+    if not df_clin is None:
+        df_clin = df_clin.loc[patients_shared, :].sort_index()
+    return df_omics, df_label, df_clin
+
+
+# api
+def read_dataset(args):
+    # 1. read raw dataset
+    # (1) read omics dataset
+    df_omics = read_omics(args)
+    # (2) read label
+    df_label = read_label(args)
+    # (3) read clinical feature
+    df_clin = read_clin(args)
+
+    # 2. process
+    df_omics, df_label, df_clin = process(df_omics, df_label, df_clin)
+
+    # 3. return clean dataset
+    return df_omics, df_label, df_clin