DC平台蛋白质分子匹配预测.zip

import pandas as pd import numpy as np import gc import re import sys import time import jieba import os.path import os import datetime import numpy as np import pandas as pd import lightgbm as lgb import warnings warnings.filterwarnings(action='ignore', category=UserWarning, module='gensim') import gensim from gensim.models import Word2Vec df_molecule=pd.read_csv('df_molecule.csv') df_protein_test_data=pd.read_csv('df_protein_test.csv') df_affinity_test_toBePredicted=pd.read_csv('df_affinity_test_toBePredicted.csv') df_protein_train_data=pd.read_csv('df_protein_train.csv') df_affinity_train=pd.read_csv('df_affinity_train.csv') #用于保存训练集标签 train_label=df_affinity_train['Ki'] #train_label=float(train_label) del df_affinity_train['Ki'] #提取蛋白质序列基本数量特征 def fun_D_N_rate(x): m=x.count('D')+x.count('N') a=float(m/len(x)) return a def fun_ALL_rate(x): m = x.count('ALL') a = float(m / len(x)) return a def fun_water_rate(x): m = x.count('V')+x.count('I')+x.count('L')+x.count('F')+x.count('W')+x.count('Y')+x.count('M') a = float(m / len(x)) return a def fun_f_w_y_h_rate(x): m = x.count('H')+ x.count('F') + x.count('W') + x.count('Y') a = float(m / len(x)) return a def fun_v_i_l_m_rate(x): m = x.count('V')+ x.count('I') + x.count('L') + x.count('M') a = float(m / len(x)) return a def fun_lovewater_rate(x): m = x.count('S')+x.count('T')+x.count('H')+x.count('N')+x.count('Q')+x.count('E')+x.count('D')+x.count('K')+x.count('R') a = float(m / len(x)) return a def fun_krh_rate(x): m = x.count('K')+ x.count('R') + x.count('H') a = float(m / len(x)) return a def fun_de_rate(x): m = x.count('D')+ x.count('E') a = float(m / len(x)) return a def fun_PGAS_rate(x): m = x.count('P')+ x.count('G') + x.count('A') + x.count('S') a = float(m / len(x)) return a def fun_zhengfulizi_rate(x): m2 = x.count('K') + x.count('R') + x.count('H') m1 = x.count('D') + x.count('E') try: a = float(m2 / m1) return a except: return 0 def water_likeorhate_rate(x): m1 = x.count('S') + x.count('T') + x.count('H') + x.count('N') + x.count('Q') + x.count('E') + x.count( 'D') + x.count('K') + x.count('R') m2 = x.count('V') + x.count('I') + x.count('L') + x.count('F') + x.count('W') + x.count('Y') + x.count('M') try: a=float(m1/m2) return a except: return 0 def protein_sequence_num_feat(df): df['A_num']=df.Sequence.map(lambda x:x.count('A')) df['R_num']=df.Sequence.map(lambda x:x.count('R')) df['N_num'] = df.Sequence.map(lambda x: x.count('N')) df['D_num'] = df.Sequence.map(lambda x: x.count('D')) df['C_num'] = df.Sequence.map(lambda x: x.count('C')) df['E_num'] = df.Sequence.map(lambda x: x.count('E')) df['Q_num'] = df.Sequence.map(lambda x: x.count('Q')) df['G_num'] = df.Sequence.map(lambda x: x.count('G')) df['H_num'] = df.Sequence.map(lambda x: x.count('H')) df['I_num'] = df.Sequence.map(lambda x: x.count('I')) df['L_num'] = df.Sequence.map(lambda x: x.count('L')) df['K_num'] = df.Sequence.map(lambda x: x.count('K')) df['M_num'] = df.Sequence.map(lambda x: x.count('M')) df['F_num'] = df.Sequence.map(lambda x: x.count('F')) df['P_num'] = df.Sequence.map(lambda x: x.count('P')) df['S_num'] = df.Sequence.map(lambda x: x.count('S')) df['T_num'] = df.Sequence.map(lambda x: x.count('T')) df['W_num'] = df.Sequence.map(lambda x: x.count('W')) df['Y_num'] = df.Sequence.map(lambda x: x.count('Y')) df['V_num'] = df.Sequence.map(lambda x: x.count('V')) df['len_se']=df.Sequence.map(lambda x: len(x)) #df['D_N_rate']=df.Sequence.map(lambda x: fun_D_N_rate(x)) #df['all_rate'] = df.Sequence.map(lambda x: fun_ALL_rate(x)) #df['water_rate']=df.Sequence.map(lambda x: fun_water_rate(x)) #df['f_w_y_h_rate']=df.Sequence.map(lambda x: fun_f_w_y_h_rate(x)) #df['v_i_l_m_rate']=df.Sequence.map(lambda x: fun_v_i_l_m_rate(x)) #df['love_water']=df.Sequence.map(lambda x: fun_lovewater_rate(x)) #df['k_r_h_rate']=df.Sequence.map(lambda x: fun_krh_rate(x)) #df['d_e_rate']=df.Sequence.map(lambda x: fun_de_rate(x)) #df['pgas_rate']=df.Sequence.map(lambda x: fun_PGAS_rate(x)) #df['lizi_rate']=df.Sequence.map(lambda x: fun_zhengfulizi_rate(x)) #df['water_likeorhate_rate']=df.Sequence.map(lambda x: water_likeorhate_rate(x)) return df #构建训练和测试样本的蛋白质特征 df_protein_train_feat=protein_sequence_num_feat(df_protein_train_data) #del df_protein_train_feat['Sequence'] df_protein_test_feat=protein_sequence_num_feat(df_protein_test_data) #del df_protein_test_feat['Sequence'] i=0 while(i<len(df_molecule['Fingerprint'][1])): df_molecule['Fingerprint'+'_'+str(i)]=df_molecule.Fingerprint.map(lambda x:int(x[i])) i=i+3 df_protein_train_feat=pd.merge(df_affinity_train,df_protein_train_feat,on='Protein_ID',how='left') del df_molecule['Fingerprint'] df_protein_train_feat=pd.merge(df_protein_train_feat,df_molecule,on='Molecule_ID',how='left') df_protein_train_feat=df_protein_train_feat.fillna(df_protein_train_feat.mean()) df_protein_test_feat=pd.merge(df_affinity_test_toBePredicted,df_protein_test_feat,on='Protein_ID',how='left') df_protein_test_feat=pd.merge(df_protein_test_feat,df_molecule,on='Molecule_ID',how='left') df_protein_test_feat=df_protein_test_feat.fillna(df_protein_test_feat.mean()) #test_vector n = 128 texts = [[word for word in re.findall(r'.{3}',document)] for document in list(df_protein_test_data['Sequence'])] model = Word2Vec(texts,size=n,window=4,min_count=1,negative=3, sg=1,sample=0.001,hs=1,workers=4) vectors = pd.DataFrame([model[word] for word in (model.wv.vocab)]) vectors['Word'] = list(model.wv.vocab) vectors.columns= ["vec_{0}".format(i) for i in range(0,n)]+["Word"] #train_vector texts2 = [[word for word in re.findall(r'.{3}',document)] for document in list(df_protein_train_data['Sequence'])] model2 = Word2Vec(texts,size=n,window=4,min_count=1,negative=3, sg=1,sample=0.001,hs=1,workers=4) vectors2 = pd.DataFrame([model[word] for word in (model.wv.vocab)]) vectors2['Word'] = list(model.wv.vocab) vectors2.columns= ["vec_{0}".format(i) for i in range(0,n)]+["Word"] vectors['Protein_ID']=df_protein_test_data['Protein_ID'] vectors2['Protein_ID']=df_protein_train_data['Protein_ID'] del vectors2['Word'] del vectors['Word'] print('vector_get_finish') df_protein_train_feat=pd.merge(df_protein_train_feat,vectors2,on='Protein_ID',how='left') df_protein_test_feat=pd.merge(df_protein_test_feat,vectors,on='Protein_ID',how='left') del df_protein_test_feat['Protein_ID'] del df_protein_test_feat['Molecule_ID'] del df_protein_train_feat['Protein_ID'] del df_protein_train_feat['Molecule_ID'] del df_protein_train_feat['Sequence'] del df_protein_test_feat['Sequence'] print('feature_finish') #---------------------------------------------------------------------------- from sklearn.model_selection import train_test_split x_train,x_val,y_train,y_val=train_test_split(df_protein_train_feat,train_label,test_size=0.2,random_state=100) import xgboost as xgb print ('start running ....') dtrain = xgb.DMatrix(x_train,label=y_train) dval = xgb.DMatrix(x_val,label=y_val) param = {'learning_rate' : 0.1, 'n_estimators': 1000, 'max_depth': 4, 'min_child_weight': 6, 'gamma': 0, 'subsample': 0.8, 'colsample_bytree': 0.8, 'eta': 0.05, 'silent': 1, } num_round =150 plst = list(param.items()) plst += [('eval_metric', 'rmse')] evallist = [(dval, 'eval'), (dtrain, 'train')] bst=xgb.train(plst,dtrain,num_round,evallist,early_stopping_rounds=10) dtest = xgb.DMatrix(df_protein_test_feat) Pred = bst.predict(dtest) df_affinity_test_toBePredicted['Ki']=Pred df_affinity_test_toBePredicted.to_csv('baseline_4_12.csv',index=False)