File size: 7,220 Bytes
90d94ec | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 | import _bootstrap
from collections import Counter
from scipy.ndimage import convolve1d
from scipy.ndimage import gaussian_filter1d
from scipy.signal.windows import triang
import matplotlib.pyplot as plt
import torch
from model.build_vocab import WordVocab
from model.pretrain_trfm import TrfmSeq2seq
from model.utils import split
import json
from transformers import T5EncoderModel, T5Tokenizer
import re
import gc
from sklearn import metrics
from sklearn.ensemble import ExtraTreesRegressor
import numpy as np
import pandas as pd
from sklearn.model_selection import KFold
from sklearn.metrics import mean_squared_error, mean_absolute_error
from sklearn.metrics import r2_score
from sklearn.model_selection import train_test_split
import random
import pickle
import math
from project_paths import (
DEGREE_SMILES_PATH, KCAT_DATASET_PATH, KCAT_KM_SAMPLES_PATH, KM_TEST_PATH,
PH_SMILES_PATH, PROT_T5_MODEL, TRFM_PATH, UNIKP_MODEL_DIR, VOCAB_PATH,
)
def smiles_to_vec(Smiles):
pad_index = 0
unk_index = 1
eos_index = 2
sos_index = 3
mask_index = 4
vocab = WordVocab.load_vocab(VOCAB_PATH)
def get_inputs(sm):
seq_len = 220
sm = sm.split()
if len(sm)>218:
print('SMILES is too long ({:d})'.format(len(sm)))
sm = sm[:109]+sm[-109:]
ids = [vocab.stoi.get(token, unk_index) for token in sm]
ids = [sos_index] + ids + [eos_index]
seg = [1]*len(ids)
padding = [pad_index]*(seq_len - len(ids))
ids.extend(padding), seg.extend(padding)
return ids, seg
def get_array(smiles):
x_id, x_seg = [], []
for sm in smiles:
a,b = get_inputs(sm)
x_id.append(a)
x_seg.append(b)
return torch.tensor(x_id), torch.tensor(x_seg)
trfm = TrfmSeq2seq(len(vocab), 256, len(vocab), 4)
trfm.load_state_dict(torch.load(TRFM_PATH))
trfm.eval()
x_split = [split(sm) for sm in Smiles]
xid, xseg = get_array(x_split)
X = trfm.encode(torch.t(xid))
return X
def Seq_to_vec(Sequence):
for i in range(len(Sequence)):
if len(Sequence[i]) > 1000:
Sequence[i] = Sequence[i][:500] + Sequence[i][-500:]
sequences_Example = []
for i in range(len(Sequence)):
zj = ''
for j in range(len(Sequence[i]) - 1):
zj += Sequence[i][j] + ' '
zj += Sequence[i][-1]
sequences_Example.append(zj)
tokenizer = T5Tokenizer.from_pretrained(PROT_T5_MODEL, do_lower_case=False)
model = T5EncoderModel.from_pretrained(PROT_T5_MODEL)
gc.collect()
print(torch.cuda.is_available())
# 'cuda:0' if torch.cuda.is_available() else
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
model = model.to(device)
model = model.eval()
features = []
for i in range(len(sequences_Example)):
print('For sequence ', str(i+1))
sequences_Example_i = sequences_Example[i]
sequences_Example_i = [re.sub(r"[UZOB]", "X", sequences_Example_i)]
ids = tokenizer.batch_encode_plus(sequences_Example_i, add_special_tokens=True, padding=True)
input_ids = torch.tensor(ids['input_ids']).to(device)
attention_mask = torch.tensor(ids['attention_mask']).to(device)
with torch.no_grad():
embedding = model(input_ids=input_ids, attention_mask=attention_mask)
embedding = embedding.last_hidden_state.cpu().numpy()
for seq_num in range(len(embedding)):
seq_len = (attention_mask[seq_num] == 1).sum()
seq_emd = embedding[seq_num][:seq_len - 1]
features.append(seq_emd)
features_normalize = np.zeros([len(features), len(features[0][0])], dtype=float)
for i in range(len(features)):
for k in range(len(features[0][0])):
for j in range(len(features[i])):
features_normalize[i][k] += features[i][j][k]
features_normalize[i][k] /= len(features[i])
return features_normalize
def Kcat_predict(Ifeature, Label, weights):
kf = KFold(n_splits=5, shuffle=True)
All_pre_label = []
All_real_label = []
for train_index, test_index in kf.split(Ifeature, Label):
Train_data, Train_label = Ifeature[train_index], Label[train_index]
Test_data, Test_label = Ifeature[test_index], Label[test_index]
model = ExtraTreesRegressor()
# , sample_weight=weights[train_index]
model.fit(Train_data, Train_label, sample_weight=weights[train_index])
Pre_label = model.predict(Test_data)
All_pre_label.extend(Pre_label)
All_real_label.extend(Test_label)
res = pd.DataFrame({'Value': All_real_label, 'Predict_Label': All_pre_label})
res.to_excel('DMW/DMW_No_Normalize_2_LDS_Kcat_5_cv'+'.xlsx')
if __name__ == '__main__':
# Dataset Load
with open(KCAT_DATASET_PATH, 'r') as file:
datasets = json.load(file)
# print(len(datasets))
# datasets = datasets[:50]
sequence = [data['Sequence'] for data in datasets]
Smiles = [data['Smiles'] for data in datasets]
Label = [float(data['Value']) for data in datasets]
ECNumber = [data['ECNumber'] for data in datasets]
Organism = [data['Organism'] for data in datasets]
Substrate = [data['Substrate'] for data in datasets]
Type = [data['Type'] for data in datasets]
for i in range(len(Label)):
if Label[i] == 0:
Label[i] = -10000000000
else:
Label[i] = math.log(Label[i], 10)
Label = np.array(Label)
print(max(Label), min(Label))
# Feature Extractor
# smiles_input = smiles_to_vec(Smiles)
# sequence_input = Seq_to_vec(sequence)
# feature = np.concatenate((smiles_input, sequence_input), axis=1)
with open("PreKcat_new/features_17010_PreKcat.pkl", "rb") as f:
feature = pickle.load(f)
# Input dataset
feature_new = []
Label_new = []
sequence_new = []
Smiles_new = []
ECNumber_new = []
Organism_new = []
Substrate_new = []
Type_new = []
for i in range(len(Label)):
if -10000000000 < Label[i] and '.' not in Smiles[i]:
feature_new.append(feature[i])
Label_new.append(Label[i])
sequence_new.append(sequence[i])
Smiles_new.append(Smiles[i])
ECNumber_new.append(ECNumber[i])
Organism_new.append(Organism[i])
Substrate_new.append(Substrate[i])
Type_new.append(Type[i])
print(len(Label_new), min(Label_new), max(Label_new))
feature_new = np.array(feature_new)
Label_new = np.array(Label_new)
# sl_label = [Label_new[i] for i in range(len(Label_new))]
# weights = Smooth_Label(sl_label)
weights = np.ones([len(Label_new)], dtype=float)
for i in range(len(weights)):
if Label_new[i] > 4:
weights[i] = 2
# sum_weights = np.sum(weights)
# for i in range(len(weights)):
# weights[i] /= sum_weights
# Modelling
Kcat_predict(feature_new, Label_new, weights)
|