def predict_RI(smiles, mode='SimiStdNP'):
words = open('DeepEI/data/words.json', 'r').read()
words = json.loads(words)
if mode == 'SimiStdNP':
model = load_model('Retention/models/SimiStdNP_CNN_multi_model.h5')
elif mode == 'StdNP':
model = load_model('Retention/models/StdNP_CNN_multi_model.h5')
elif mode == 'StdPolar':
model = load_model('Retention/models/StdPolar_CNN_multi_model.h5')
else:
return None
X = []
for i, smi in enumerate(smiles):
xi = one_hot_coding(smi, words, max_len=100)
X.append(xi.todense())
X = np.array(X)
pred = model.predict(X)
return pred
# -*- coding: utf-8 -*-
"""
Created on Wed Jul 3 10:20:36 2019
@author: hcji
"""
import os
import json
from tqdm import tqdm
from smiles_to_onehot.encoding import get_dict, one_hot_coding
with open('hmdb_smiles/hmdb_smiles.json', 'r') as js:
hmdb_smiles = json.load(js)
words = get_dict(hmdb_smiles, save_path=os.getcwd())
x = []
for smi in tqdm(hmdb_smiles):
xi = one_hot_coding(smi, words, max_len=100)
x.append(xi)
K.clear_session()
if __name__ == '__main__':
import json
with open('DeepEI/data/split.json', 'r') as js:
keep = np.array(json.load(js)['keep'])
smiles = np.array(json.load(open('DeepEI/data/all_smiles.json')))[keep]
rindex = np.load('DeepEI/data/retention.npy')[keep, :]
words = get_dict(smiles, save_path='DeepEI/data/words.json')
smiles = [
one_hot_coding(smi, words, max_len=100).todense() for smi in smiles
]
smiles = np.array(smiles)
# simipolar
i = np.where(~np.isnan(rindex[:, 0]))[0]
mod = multi_CNN(smiles[i], rindex[i, 0])
mod.train()
mod.test()
mod.save('Retention/models/SimiStdNP_CNN_multi_model.h5')
# nonpolar
i = np.where(~np.isnan(rindex[:, 1]))[0]
mod = multi_CNN(smiles[i], rindex[i, 1])
mod.train()
mod.test()