File size: 6,017 Bytes
b3c2eb7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from rdkit import Chem
import codecs
import numpy as np
import pandas as pd
from rdkit.Chem import Descriptors
from subword_nmt.apply_bpe import BPE
import deepchem as dc
import deepchem.molnet as dcm
from rdkit import RDLogger
from torch.utils.data import DataLoader, Dataset, Subset
from pytorch_lightning import LightningDataModule


lg = RDLogger.logger()
lg.setLevel(RDLogger.CRITICAL)


def smiles2index(s1, words2idx, bpe):
    t1 = bpe.process_line(s1).split()
    i1 = [words2idx[i] for i in t1]
    return i1


def index2multi_hot(i1, idx2word):
    v1 = np.zeros(len(idx2word))
    v1[i1] = 1
    return v1


def index2multi_hot_fg(molecule, fgroups_list):
    v1 = np.zeros(len(fgroups_list))
    for idx in range(len(fgroups_list)):
        if molecule.HasSubstructMatch(fgroups_list[idx]):
            v1[idx] = 1
    return v1


def smiles2vector_fgr(s1, words2idx, bpe, idx2word, fgroups_list):
    i1 = smiles2index(s1, words2idx, bpe)
    mfg = index2multi_hot(i1, idx2word)
    molecule = Chem.MolFromSmiles(s1)
    fg = index2multi_hot_fg(molecule, fgroups_list)
    return fg, mfg


class FsrFgDataset(Dataset):

    def __init__(self, data, idx2word, fgroups_list, words2idx, bpe, descriptor_funcs):
        self.mols = data.X
        self.y = data.y
        self.smiles = data.ids
        self.idx2word = idx2word
        self.fgroups_list = fgroups_list
        self.words2idx = words2idx
        self.bpe = bpe
        self.descriptor_funcs = descriptor_funcs

    def __len__(self):
        return len(self.y)

    def __getitem__(self, idx):
        mol = self.mols[idx]
        smile = self.smiles[idx]
        target = self.y[idx]

        fg, mfg = smiles2vector_fgr(smile, self.words2idx, self.bpe, self.idx2word, self.fgroups_list)
        num_features = np.asarray([self.descriptor_funcs[key](mol) for key in self.descriptor_funcs.keys()])
        return np.float32(fg), np.float32(mfg), np.float32(num_features), int(target)


class FsrFgDataModule(LightningDataModule):
    def __init__(self, root: str = 'Data', task_name: str = 'ecoli', batch_size: int = 8, num_workers: int = 8,
                 pin_memory: bool = True, split_type: str = 'scaffold', num_folds: int = 5, fold_index: int = 0):
        super().__init__()

        self.root = root
        self.task_name = task_name
        self.num_workers = num_workers
        self.pin_memory = pin_memory
        self.batch_size = batch_size
        self.split_type = split_type
        self.num_folds = num_folds
        self.fold_index = fold_index
        self.test_fold = None
        self.val_fold = None
        self.train_fold = None
        self.fgroups_list = None
        self.descriptor_funcs = None
        self.bpe = None
        self.idx2word = None
        self.words2idx = None
        self.val_data = None
        self.test_data = None
        self.train_data = None
        self.splits = None

    def prepare_data(self) -> None:
        df = pd.read_csv('Data/ecoli_inhibition.csv')
        ids = df['smiles']
        labels = df['Encoded_Activity']
        X = [Chem.MolFromSmiles(smiles) for smiles in ids]
        data = dc.data.DiskDataset.from_numpy(X=X, y=labels, ids=ids)
        fgroups = pd.read_csv(self.root + '/Functional_groups_filtered.csv')
        fgroups_list = list(map(lambda x: Chem.MolFromSmarts(x), fgroups['SMARTS'].tolist()))
        self.fgroups_list = [i for i in fgroups_list if i]

        self.descriptor_funcs = {name: func for name, func in Descriptors.descList}

        vocab_path = self.root + '/codes_drug_chembl_1500.txt'
        bpe_codes_fin = codecs.open(vocab_path)
        self.bpe = BPE(bpe_codes_fin, merges=-1, separator='')
        vocab_map = pd.read_csv(self.root + '/subword_units_map_drug_chembl_1500.csv')
        self.idx2word = vocab_map['index'].values
        self.words2idx = dict(zip(self.idx2word, range(0, len(self.idx2word))))
        self.weights = data.w
        if self.split_type == 'scaffold':
            splitter = dc.splits.ScaffoldSplitter()
            self.splits = []
            for i in range(self.num_folds):
                new_data = data.complete_shuffle()
                self.splits.append((new_data, splitter.split(dataset=new_data, seed=i)))
            self.dataset = FsrFgDataset(self.splits[self.fold_index][0], self.idx2word, self.fgroups_list, self.words2idx, self.bpe,
                                        self.descriptor_funcs)
            self.train_ind, self.val_ind, self.test_ind = self.splits[self.fold_index][1]
        else:
            splitter = dc.splits.RandomStratifiedSplitter()
            self.splits = [splitter.split(dataset=data, seed=fold_num) for fold_num in range(self.num_folds)]
            self.dataset = FsrFgDataset(data, self.idx2word, self.fgroups_list, self.words2idx, self.bpe,
                                        self.descriptor_funcs)
            self.train_ind, self.val_ind, self.test_ind = self.splits[self.fold_index]

    def setup(self, stage=None):
        """define train, test and validation datasets """

        self.train_fold = Subset(self.dataset, self.train_ind)
        self.val_fold = Subset(self.dataset, self.val_ind)
        self.test_fold = Subset(self.dataset, self.test_ind)

    def train_dataloader(self):
        """returns train dataloader"""
        loader = DataLoader(self.train_fold, batch_size=self.batch_size, shuffle=False,
                            num_workers=self.num_workers, pin_memory=self.pin_memory, drop_last=True)
        return loader

    def val_dataloader(self):
        """returns val dataloader"""
        loader = DataLoader(self.val_fold, batch_size=1, shuffle=False,
                            num_workers=self.num_workers, pin_memory=self.pin_memory)
        return loader

    def test_dataloader(self):
        """returns test dataloader"""
        loader = DataLoader(self.test_fold, batch_size=1, shuffle=False,
                            num_workers=self.num_workers, pin_memory=self.pin_memory)
        return loader