欢迎关注我的CSDN：https://spike.blog.csdn.net/
本文地址：https://blog.csdn.net/caroline_wendy/article/details/131201456

RoseTTAFold2 是蛋白质结构预测算法，利用了深度学习和三维几何建模的技术，能够快速准确地预测蛋白质的三级结构。RoseTTAFold2 是在 RoseTTAFold 的基础上改进的，增加了对多肽段的处理，提高了对复杂蛋白质的预测能力，同时也降低了计算资源的需求。

Paper: Efficient and accurate prediction of protein structure using RoseTTAFold2，2023.5.25

预测结果的数据路径：output/rf2-monomer/，56个样本

以T1104-D1_A117为例，具体格式如下：

hhblits/
log/
models/
T1104-D1_A117_1.atab*
T1104-D1_A117_1.fa*
T1104-D1_A117_1.hhr*
T1104-D1_A117_1.msa0.a3m*

PDB模型，位于models，PDB数据文件名，即model_00_pred.pdb。

model_00.json
model_00.npz
model_00_pred.pdb

结果导出脚本：main_rf2_results_parser.py

python scripts/main_rf2_results_parser.py -i output/rf2-monomer/ -o mydata/results_v2/casp15-fasta-56-outputs-rf2-results

源码如下：

#!/usr/bin/env python
# -- coding: utf-8 --
"""
Copyright (c) 2022. All rights reserved.
Created by C. L. Wang on 2023/6/14
"""
import argparse
import collections
import json
import os
import shutil
import sys
from pathlib import Path

from tqdm import tqdm

p = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
if p not in sys.path:
    sys.path.append(p)

from protein_utils.pdb_utils import get_plddt_from_pdb
from myutils.project_utils import traverse_dir_files, mkdir_if_not_exist


class MainRf2ResultsParser(object):
  """
  RoseTTAFold2 的结果处理，以利于后续评估
  """
  def __init__(self):
    pass

  def process(self, input_dir, output_dir):
    print(f"[Info] 输入文件夹: {input_dir}")
    print(f"[Info] 输出文件夹: {output_dir}")
    mkdir_if_not_exist(output_dir)
    path_list = traverse_dir_files(input_dir, ext="pdb")
    print(f"[Info] 样本数量: {len(path_list)}")

    fasta_dict = collections.defaultdict(list)
    for path in tqdm(path_list, desc="[Info] PDB"):
      tmp_items = path.split("/")
      fasta_name = tmp_items[-3]
      fasta_dict[fasta_name].append(path)
    print(f"[Info] FASTA数量: {len(fasta_dict.keys())}")

    for fasta_name in tqdm(fasta_dict.keys(), desc="[Info] FASTA"):
      fasta_path_list = fasta_dict[fasta_name]

      # 构建文件夹
      out_fasta_dir = os.path.join(output_dir, fasta_name)
      mkdir_if_not_exist(out_fasta_dir)

      pdb_score_list = []
      for path in fasta_path_list:
        file_name = os.path.basename(path)

        # 复制文件
        shutil.copy(path, os.path.join(out_fasta_dir, file_name))

        # 写入文件
        plddt_val = float(get_plddt_from_pdb(path))
        pdb_name = file_name.split(".")[0]
        pdb_score_list.append((pdb_name, plddt_val))

      # 写入评估结果的JSON文件
      pdb_score_list = sorted(pdb_score_list, key=lambda x: x[1], reverse=True)
      score_map_path = os.path.join(out_fasta_dir, "ranking_debug.json")
      with open(score_map_path, 'w') as f:
        score_map_dict = {
          "plddts": dict(pdb_score_list),
          "order": [x[0] for x in pdb_score_list],
        }
        json.dump(score_map_dict, f, indent=4, sort_keys=True)


def main():
  parser = argparse.ArgumentParser()
  parser.add_argument(
    "-i",
    "--input_dir",
    help="the dir from rosettafold2 output.",
    type=Path,
    required=True,
  )
  parser.add_argument(
    "-o",
    "--output-dir",
    help="output dir path.",
    type=Path,
    required=True
  )

  args = parser.parse_args()

  input_dir = str(args.input_dir)
  output_dir = str(args.output_dir)
  mkdir_if_not_exist(output_dir)

  assert os.path.isdir(input_dir) and os.path.isdir(output_dir)

  rrp = MainRf2ResultsParser()
  rrp.process(input_dir, output_dir)


if __name__ == '__main__':
  main()