Chemistry.AI|基于卷积神经网络（CNN）预测分子特性_Chemistry.AI|基于循环神经网络（RNN）预测分子性质资源-CSDN文库

194 浏览量 2020-12-21 04:38:07 上传评论 1 收藏 55KB PDF 举报

资源详情

资源评论

资源推荐

Chemistry.AI | 基于卷积神经网络（基于卷积神经网络（CNN）预测分子特性）预测分子特性

CNN :Convolutional Neural Networks (卷积神经网络 )

环境准备环境准备

Python版本：Python 3.6.8

PyTorch版本：PyTorch1.1.0

RDKit版本：RDKit 2020.03.1

基于卷积神经网络（基于卷积神经网络（CNN）预测分子特性）预测分子特性

导入库

from rdkit import Chem

from rdkit.Chem.Crippen import MolLogP

import numpy as np

import torch

import time

载入数据

maxlen = 64

with open('smiles.txt') as f:

smiles = f.readlines()[:] smiles = [s.strip() for s in smiles] smiles = [s.split()[1] for s in smiles] smiles = [s for s in smiles if len(s)<maxlen]

#Characters of smiles

all_smiles=''

for s in smiles: all_smiles+=s

chars = sorted(list(set(list(all_smiles))))

chars.append('X')

c_to_i = {c:i for i,c in enumerate(chars)}

print ('Max len:', maxlen)

print ('Number of chars:', len(chars))

print (chars)

print (c_to_i)

Max len: 64

Number of chars: 46

['#', '(', ')', '+', '-', '.', '/', '1', '2', '3', '4', '5', '6', '7', '=', '@', 'B', 'C', 'F', 'G', 'H', 'I', 'K', 'L', 'M', 'N', 'O', 'P', 'S', 'V', 'Z', '[', '\', ']', 'a', 'c', 'e', 'g', 'i', 'l', 'n', 'o', 'r', 's', 'u', 'X']

{'#': 0, '(': 1, ')': 2, '+': 3, '-': 4, '.': 5, '/': 6, '1': 7, '2': 8, '3': 9, '4': 10, '5': 11, '6': 12, '7': 13, '=': 14, '@': 15, 'B': 16, 'C': 17, 'F': 18, 'G': 19, 'H': 20, 'I': 21, 'K': 22,

'L': 23, 'M': 24, 'N': 25, 'O': 26, 'P': 27, 'S': 28, 'V': 29, 'Z': 30, '[': 31, '\': 32, ']': 33, 'a': 34, 'c': 35, 'e': 36, 'g': 37, 'i': 38, 'l': 39, 'n': 40, 'o': 41, 'r': 42, 's': 43, 'u':

44, 'X': 45}

计算每个分子的分子指纹和LogP

Y = [] num_data = 20000

st = time.time()

for s in smiles[:num_data]:

m = Chem.MolFromSmiles(s)

logp = MolLogP(m)

Y.append(logp)

end = time.time()

print (f'Time:{(end-st):.3f}')

数据批处理

from torch.utils.data import Dataset, DataLoader