File size: 4,957 Bytes
48b5986 | 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 | from pathlib import Path
import sys
_PROJECT_FILE = Path(__file__).resolve()
for _PROJECT_ROOT in _PROJECT_FILE.parents:
if (_PROJECT_ROOT / "model").is_dir():
if str(_PROJECT_ROOT) not in sys.path:
sys.path.insert(0, str(_PROJECT_ROOT))
break
# Copyright (c) Meta Platforms, Inc. and affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
#
# Scores sequences based on a given structure.
#
# usage:
# score_log_likelihoods.py [-h] [--outpath OUTPATH] [--chain CHAIN] pdbfile seqfile
import argparse
from biotite.sequence.io.fasta import FastaFile, get_sequences
import numpy as np
from pathlib import Path
import torch
import torch.nn.functional as F
from tqdm import tqdm
import model.esm as esm
import model.esm.inverse_folding
def score_singlechain_backbone(model, alphabet, args):
if torch.cuda.is_available() and not args.nogpu:
model = model.cuda()
print("Transferred model to GPU")
coords, native_seq = esm.inverse_folding.util.load_coords(args.pdbfile, args.chain)
print('Native sequence loaded from structure file:')
print(native_seq)
print('\n')
ll, _ = esm.inverse_folding.util.score_sequence(
model, alphabet, coords, native_seq)
print('Native sequence')
print(f'Log likelihood: {ll:.2f}')
print(f'Perplexity: {np.exp(-ll):.2f}')
print('\nScoring variant sequences from sequence file..\n')
infile = FastaFile()
infile.read(args.seqfile)
seqs = get_sequences(infile)
Path(args.outpath).parent.mkdir(parents=True, exist_ok=True)
with open(args.outpath, 'w') as fout:
fout.write('seqid,log_likelihood\n')
for header, seq in tqdm(seqs.items()):
ll, _ = esm.inverse_folding.util.score_sequence(
model, alphabet, coords, str(seq))
fout.write(header + ',' + str(ll) + '\n')
print(f'Results saved to {args.outpath}')
def score_multichain_backbone(model, alphabet, args):
if torch.cuda.is_available() and not args.nogpu:
model = model.cuda()
print("Transferred model to GPU")
structure = esm.inverse_folding.util.load_structure(args.pdbfile)
coords, native_seqs = esm.inverse_folding.multichain_util.extract_coords_from_complex(structure)
target_chain_id = args.chain
native_seq = native_seqs[target_chain_id]
print('Native sequence loaded from structure file:')
print(native_seq)
print('\n')
ll, _ = esm.inverse_folding.multichain_util.score_sequence_in_complex(
model, alphabet, coords, target_chain_id, native_seq)
print('Native sequence')
print(f'Log likelihood: {ll:.2f}')
print(f'Perplexity: {np.exp(-ll):.2f}')
print('\nScoring variant sequences from sequence file..\n')
infile = FastaFile()
infile.read(args.seqfile)
seqs = get_sequences(infile)
Path(args.outpath).parent.mkdir(parents=True, exist_ok=True)
with open(args.outpath, 'w') as fout:
fout.write('seqid,log_likelihood\n')
for header, seq in tqdm(seqs.items()):
ll, _ = esm.inverse_folding.multichain_util.score_sequence_in_complex(
model, alphabet, coords, target_chain_id, str(seq))
fout.write(header + ',' + str(ll) + '\n')
print(f'Results saved to {args.outpath}')
def main():
parser = argparse.ArgumentParser(
description='Score sequences based on a given structure.'
)
parser.add_argument(
'pdbfile', type=str,
help='input filepath, either .pdb or .cif',
)
parser.add_argument(
'seqfile', type=str,
help='input filepath for variant sequences in a .fasta file',
)
parser.add_argument(
'--outpath', type=str,
help='output filepath for scores of variant sequences',
default='output/sequence_scores.csv',
)
parser.add_argument(
'--chain', type=str,
help='chain id for the chain of interest', default='A',
)
parser.set_defaults(multichain_backbone=False)
parser.add_argument(
'--multichain-backbone', action='store_true',
help='use the backbones of all chains in the input for conditioning'
)
parser.add_argument(
'--singlechain-backbone', dest='multichain_backbone',
action='store_false',
help='use the backbone of only target chain in the input for conditioning'
)
parser.add_argument("--nogpu", action="store_true", help="Do not use GPU even if available")
args = parser.parse_args()
model, alphabet = esm.pretrained.esm_if1_gvp4_t16_142M_UR50()
model = model.eval()
if args.multichain_backbone:
score_multichain_backbone(model, alphabet, args)
else:
score_singlechain_backbone(model, alphabet, args)
if __name__ == '__main__':
main()
|