OpenFold2/alphafold.py at main · lupoglaz/OpenFold2 · GitHub

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
from genericpath import exists
import os
import sys
import pathlib
from typing import Dict
import argparse
import random
import time
import subprocess
import json
from pathlib import Path
import io
import numpy as np
import pickle
import torch

from alphafold.Model.alphafold import AlphaFold
from alphafold.Model import model_config
from alphafold.Model.features import AlphaFoldFeatures
from alphafold.Model.Utils.weights_loading import params_to_torch
from alphafold.Common import protein, residue_constants
from alphafold.Data import pipeline


if __name__ == '__main__':
	parser = argparse.ArgumentParser(description='Train deep protein docking')
	parser.add_argument('-fasta_path', default='T1024.fas', type=str)
	parser.add_argument('-feature_path', default='/media/lupoglaz/OpenFold2Output/T1024/features.pkl', type=str)
	parser.add_argument('-output_dir', default='/media/lupoglaz/OpenFold2Output', type=str)
	# parser.add_argument('-output_dir', default='/media/HDD/AlphaFold2Output', type=str)
	parser.add_argument('-model_name', default='model_1', type=str)
	parser.add_argument('-data_dir', default='/media/lupoglaz/AlphaFold2Data', type=str)
	# parser.add_argument('-data_dir', default='/media/HDD/AlphaFold2', type=str)

	parser.add_argument('-jackhmmer_binary_path', default='jackhmmer', type=str)
	parser.add_argument('-hhblits_binary_path', default='hhblits', type=str)
	parser.add_argument('-hhsearch_binary_path', default='hhsearch', type=str)
	parser.add_argument('-kalign_binary_path', default='kalign', type=str)

	parser.add_argument('-uniref90_database_path', default='uniref90/uniref90.fasta', type=str)
	parser.add_argument('-mgnify_database_path', default='mgnify/mgy_clusters_2018_12.fa', type=str)
	parser.add_argument('-bfd_database_path', default='bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt', type=str)
	parser.add_argument('-small_bfd_database_path', default='small_bfd/bfd-first_non_consensus_sequences.fasta', type=str)
	parser.add_argument('-uniclust30_database_path', default='uniclust30/uniclust30_2018_08/uniclust30_2018_08', type=str)
	parser.add_argument('-pdb70_database_path', default='pdb70/pdb70', type=str)
	parser.add_argument('-template_mmcif_dir', default='pdb_mmcif/mmcif_files', type=str)
	parser.add_argument('-obsolete_pdbs_path', default='pdb_mmcif/obsolete.dat', type=str)

	parser.add_argument('-max_template_date', default='2020-05-14', type=str)
	parser.add_argument('-preset', default='reduced_dbs', type=str)
	parser.add_argument('-benchmark', default=False, type=int)
	parser.add_argument('-random_seed', default=None, type=int)

	args = parser.parse_args()
	args.fasta_path = Path(args.fasta_path)
	args.output_dir = Path(args.output_dir)
	args.feature_path = Path(args.feature_path)
	args.uniref90_database_path = Path(args.data_dir)/Path(args.uniref90_database_path)
	args.mgnify_database_path = Path(args.data_dir)/Path(args.mgnify_database_path)
	args.bfd_database_path = Path(args.data_dir)/Path(args.bfd_database_path)
	args.small_bfd_database_path = Path(args.data_dir)/Path(args.small_bfd_database_path)
	args.uniclust30_database_path = Path(args.data_dir)/Path(args.uniclust30_database_path)
	args.pdb70_database_path = Path(args.data_dir)/Path(args.pdb70_database_path)
	args.template_mmcif_dir = Path(args.data_dir)/Path(args.template_mmcif_dir)

	args.jackhmmer_binary_path = Path(subprocess.run(["which", "jackhmmer"], stdout=subprocess.PIPE).stdout[:-1].decode('utf-8'))
	args.hhblits_binary_path = Path(subprocess.run(["which", "hhblits"], stdout=subprocess.PIPE).stdout[:-1].decode('utf-8'))
	args.hhsearch_binary_path = Path(subprocess.run(["which", "hhsearch"], stdout=subprocess.PIPE).stdout[:-1].decode('utf-8'))
	args.kalign_binary_path = Path(subprocess.run(["which", "kalign"], stdout=subprocess.PIPE).stdout[:-1].decode('utf-8'))

	data_pipeline = pipeline.DataPipeline(
		jackhammer_binary_path=args.jackhmmer_binary_path,
		hhblits_binary_path=args.hhblits_binary_path,
		hhsearch_binary_path=args.hhsearch_binary_path,
		uniref90_database_path=args.uniref90_database_path,
		mgnify_database_path=args.mgnify_database_path,
		bfd_database_path=args.bfd_database_path,
		uniclust30_database_path=args.uniclust30_database_path,
		small_bfd_database_path=args.small_bfd_database_path,
		pdb70_database_path=args.pdb70_database_path,
		template_featurizer=None,
		use_small_bfd=True)

	output_dir = args.output_dir / Path(args.fasta_path.stem)
	output_dir.mkdir(parents=True, exist_ok=True)
	output_msa_dir = output_dir / Path('MSA')
	output_msa_dir.mkdir(parents=True, exist_ok=True)

	if not args.feature_path.exists():
		feature_dict = data_pipeline.process(input_fasta_path=args.fasta_path,
											msa_output_dir=output_msa_dir)
		with open(output_dir / Path('features.pkl'), 'wb') as f:
			pickle.dump(feature_dict, f, protocol=4)
	else:
		with open(args.feature_path, 'rb') as f:
			feature_dict = pickle.load(f)


	config = model_config(args.model_name)
	model_config = config.model
	model_config.embeddings_and_evoformer.template.enabled = False
	model_config.resample_msa_in_recycling = False
	data_config = config.data
	data_config.eval.num_ensemble = 1
	data_config.common.use_templates = False

	af2features = AlphaFoldFeatures(config=config)
	batch = af2features(feature_dict, random_seed=42)
	for key in batch.keys():
		print(key, batch[key].shape)
		batch[key] = batch[key][0].unsqueeze(dim=0).to(device='cuda')

	af2 = AlphaFold(config=model_config, target_dim=22, msa_dim=49, extra_msa_dim=25)

	path = os.path.join(args.data_dir, 'params', f'params_{args.model_name}.npz')
	with open(path, 'rb') as f:
		params = np.load(io.BytesIO(f.read()), allow_pickle=False)
	params = params_to_torch(params)
	af2.load_weights_from_af2(params, rel_path='alphafold')
	af2 = af2.cuda()

	with torch.no_grad():
		prediction_result, _ = af2(batch, is_training=False, compute_loss=False)
		with open(output_dir / Path('result.pkl'), 'wb') as f:
			pickle.dump(prediction_result, f, protocol=4)
		# with open('result.pkl', 'rb') as f:
		# 	prediction_result = pickle.load(f)

		protein_pdb = protein.from_prediction(features=batch, result=prediction_result)
		with open(output_dir / Path('test.pdb'), 'w') as f:
			f.write(protein.to_pdb(protein_pdb))