|
| 1 | +import seaborn as sns |
| 2 | +import matplotlib.pyplot as plt |
| 3 | +import pandas as pd |
| 4 | +import os |
| 5 | +import MDAnalysis as mda |
| 6 | +from MDAnalysis.analysis import rms |
| 7 | +from ast import literal_eval |
| 8 | +import subprocess |
| 9 | +from Bio import PDB |
| 10 | +import numpy as np |
| 11 | +import argparse |
| 12 | + |
| 13 | + |
| 14 | +def analysis(args): |
| 15 | + start_idx_dict = { |
| 16 | + "1prw": [15, 51], |
| 17 | + "1bcf": [17, 46, 90, 122], |
| 18 | + "5tpn": [108], |
| 19 | + "3ixt": [0], |
| 20 | + "4jhw": [37, 144], |
| 21 | + "4zyp": [357], |
| 22 | + "5wn9": [1], |
| 23 | + "5ius": [34, 88], |
| 24 | + "5yui": [89, 114, 194], |
| 25 | + "6vw1": [5, 45], |
| 26 | + "1qjg": [13, 37, 98], |
| 27 | + "1ycr": [2], |
| 28 | + "2kl8": [0, 27], |
| 29 | + "7mrx": [25], |
| 30 | + "5trv": [45], |
| 31 | + "6e6r": [22], |
| 32 | + "6exz": [25], |
| 33 | + } |
| 34 | + end_idx_dict = { |
| 35 | + "1prw": [34, 70], |
| 36 | + "1bcf": [24, 53, 98, 129], |
| 37 | + "5tpn": [126], |
| 38 | + "3ixt": [23], |
| 39 | + "4jhw": [43, 159], |
| 40 | + "4zyp": [371], |
| 41 | + "5wn9": [20], |
| 42 | + "5ius": [53, 109], |
| 43 | + "5yui": [93, 116, 196], |
| 44 | + "6vw1": [23, 63], |
| 45 | + "1qjg": [13, 37, 98], |
| 46 | + "1ycr": [10], |
| 47 | + "2kl8": [6, 78], |
| 48 | + "7mrx": [46], |
| 49 | + "5trv": [69], |
| 50 | + "6e6r": [34], |
| 51 | + "6exz": [39], |
| 52 | + } |
| 53 | + |
| 54 | + def calculate_avg_plddt(pdb_file): |
| 55 | + # 创建PDB解析器 |
| 56 | + parser = PDB.PDBParser(QUIET=True) |
| 57 | + |
| 58 | + # 解析PDB文件 |
| 59 | + structure = parser.get_structure("protein", pdb_file) |
| 60 | + |
| 61 | + # 获取所有的plDDT值 |
| 62 | + plddt_values = [] |
| 63 | + for model in structure: |
| 64 | + for chain in model: |
| 65 | + for residue in chain: |
| 66 | + if "CA" in residue: |
| 67 | + # 获取 CA 原子的 B-factor，并假设它存储了 plDDT 值 |
| 68 | + ca_atom = residue["CA"] |
| 69 | + plddt = ca_atom.get_bfactor() |
| 70 | + plddt_values.append(plddt) |
| 71 | + |
| 72 | + # 计算平均plDDT值 |
| 73 | + if plddt_values: |
| 74 | + avg_plddt = np.mean(plddt_values) |
| 75 | + return avg_plddt |
| 76 | + else: |
| 77 | + raise NotImplementedError |
| 78 | + |
| 79 | + def calc_rmsd_tmscore( |
| 80 | + pdb_name, |
| 81 | + reference_PDB, |
| 82 | + scaffold_pdb_path=None, |
| 83 | + scaffold_info_path=None, |
| 84 | + ref_motif_starts=[30], |
| 85 | + ref_motif_ends=[44], |
| 86 | + output_path=None, |
| 87 | + ): |
| 88 | + "Calculate RMSD between reference structure and generated structure over the defined motif regions" |
| 89 | + |
| 90 | + motif_df = pd.read_csv( |
| 91 | + os.path.join(scaffold_info_path, f"{pdb_name}.csv"), index_col=0 |
| 92 | + ) # , nrows=num_structures) |
| 93 | + results = [] |
| 94 | + for pdb in os.listdir( |
| 95 | + os.path.join(scaffold_pdb_path, f"{pdb_name}") |
| 96 | + ): # This needs to be in numerical order to match new_starts file |
| 97 | + if not pdb.endswith(".pdb"): |
| 98 | + continue |
| 99 | + ref = mda.Universe(reference_PDB) |
| 100 | + predict_PDB = os.path.join( |
| 101 | + os.path.join(scaffold_pdb_path, f"{pdb_name}"), pdb |
| 102 | + ) |
| 103 | + u = mda.Universe(predict_PDB) |
| 104 | + |
| 105 | + ref_selection = "name CA and resnum " |
| 106 | + u_selection = "name CA and resnum " |
| 107 | + i = int(pdb.split("_")[1].split(".")[0]) |
| 108 | + new_motif_starts = literal_eval(motif_df["start_idxs"].iloc[i]) |
| 109 | + new_motif_ends = literal_eval(motif_df["end_idxs"].iloc[i]) |
| 110 | + |
| 111 | + for j in range(len(ref_motif_starts)): |
| 112 | + ref_selection += ( |
| 113 | + str(ref_motif_starts[j]) + ":" + str(ref_motif_ends[j]) + " " |
| 114 | + ) |
| 115 | + u_selection += ( |
| 116 | + str(new_motif_starts[j] + 1) |
| 117 | + + ":" |
| 118 | + + str(new_motif_ends[j] + 1) |
| 119 | + + " " |
| 120 | + ) |
| 121 | + print("U SELECTION", u_selection) |
| 122 | + print("SEQUENCE", i) |
| 123 | + print("ref", ref.select_atoms(ref_selection).resnames) |
| 124 | + print("gen", u.select_atoms(u_selection).resnames) |
| 125 | + # This asserts that the motif sequences are the same - if you get this error something about your indices are incorrect - check chain/numbering |
| 126 | + assert len(ref.select_atoms(ref_selection).resnames) == len( |
| 127 | + u.select_atoms(u_selection).resnames |
| 128 | + ), "Motif lengths do not match, check PDB preprocessing \ |
| 129 | + for extra residues" |
| 130 | + |
| 131 | + assert ( |
| 132 | + ref.select_atoms(ref_selection).resnames |
| 133 | + == u.select_atoms(u_selection).resnames |
| 134 | + ).all(), "Resnames for motifRMSD do not match, check indexing" |
| 135 | + rmsd = rms.rmsd( |
| 136 | + u.select_atoms(u_selection).positions, |
| 137 | + # coordinates to align |
| 138 | + ref.select_atoms(ref_selection).positions, |
| 139 | + # reference coordinates |
| 140 | + center=True, # subtract the center of geometry |
| 141 | + superposition=True, |
| 142 | + ) # superimpose coordinates |
| 143 | + |
| 144 | + temp_file = open(os.path.join(output_path, "temp_tmscores.txt"), "w") |
| 145 | + |
| 146 | + subprocess.call( |
| 147 | + ["./analysis/TMscore", reference_PDB, predict_PDB, "-seq"], |
| 148 | + stdout=temp_file, |
| 149 | + ) |
| 150 | + with open(os.path.join(output_path, "temp_tmscores.txt"), "r") as f: |
| 151 | + for line in f: |
| 152 | + if len(line.split()) > 1 and "TM-score" == line.split()[0]: |
| 153 | + tm_score = line.split()[2] |
| 154 | + break |
| 155 | + |
| 156 | + # plddt = float(predict_PDB.split('_')[-1][:-4]) |
| 157 | + # 计算平均plDDT值 |
| 158 | + plddt = calculate_avg_plddt(predict_PDB) |
| 159 | + results.append((pdb_name, i, rmsd, plddt, tm_score)) |
| 160 | + return results |
| 161 | + |
| 162 | + scaffold_dir = args.scaffold_dir |
| 163 | + output_dir = os.path.join(scaffold_dir, "scaffold_results") |
| 164 | + os.makedirs(output_dir, exist_ok=True) |
| 165 | + |
| 166 | + results = [] |
| 167 | + for pdb in start_idx_dict.keys(): |
| 168 | + print(pdb) |
| 169 | + ref_motif_starts = start_idx_dict[pdb] |
| 170 | + ref_motif_ends = end_idx_dict[pdb] |
| 171 | + reference_PDB = os.path.join( |
| 172 | + "./data-bin/scaffolding-pdbs", pdb + "_reference.pdb" |
| 173 | + ) |
| 174 | + with open(reference_PDB) as f: |
| 175 | + line = f.readline() |
| 176 | + ref_basenum = int(line.split()[5]) |
| 177 | + ref_motif_starts = [num + ref_basenum for num in ref_motif_starts] |
| 178 | + ref_motif_ends = [num + ref_basenum for num in ref_motif_ends] |
| 179 | + results += calc_rmsd_tmscore( |
| 180 | + pdb_name=pdb, |
| 181 | + reference_PDB=reference_PDB, |
| 182 | + scaffold_pdb_path=f"{scaffold_dir}/scaffold_fasta/esmfold_pdb", |
| 183 | + scaffold_info_path=f"{scaffold_dir}/scaffold_info", |
| 184 | + ref_motif_starts=ref_motif_starts, |
| 185 | + ref_motif_ends=ref_motif_ends, |
| 186 | + output_path=output_dir, |
| 187 | + ) |
| 188 | + |
| 189 | + results = pd.DataFrame( |
| 190 | + results, columns=["pdb_name", "index", "rmsd", "plddt", "tmscore"] |
| 191 | + ) |
| 192 | + results.to_csv(os.path.join(output_dir, "rmsd_tmscore.csv"), index=False) |
| 193 | + |
| 194 | + |
| 195 | +def cal_success_scaffold(pdb): |
| 196 | + total = len(pdb) |
| 197 | + pdb["total"] = total |
| 198 | + pdb = pdb[(pdb["rmsd"] < 1.0) & (pdb["plddt"] > 70)] |
| 199 | + return pdb |
| 200 | + |
| 201 | + |
| 202 | +def motif_evaluation(args): |
| 203 | + analysis(args) |
| 204 | + |
| 205 | + output_dir = os.path.join(args.scaffold_dir, "scaffold_results") |
| 206 | + rmsd_tmscore = pd.read_csv(os.path.join(output_dir, "rmsd_tmscore.csv")) |
| 207 | + success_scaffold = rmsd_tmscore.groupby("pdb_name", as_index=False).apply( |
| 208 | + cal_success_scaffold |
| 209 | + ) |
| 210 | + success_scaffold_count = success_scaffold.groupby("pdb_name").size() |
| 211 | + success_scaffold_count = success_scaffold_count.reset_index(name="success_count") |
| 212 | + |
| 213 | + all_pdb = list(rmsd_tmscore["pdb_name"].unique()) |
| 214 | + success_pdb = list(success_scaffold_count["pdb_name"]) |
| 215 | + failed_pdb = list(set(all_pdb) - set(success_pdb)) |
| 216 | + failed_scaffold_count = { |
| 217 | + "pdb_name": failed_pdb, |
| 218 | + "success_count": [0] * len(failed_pdb), |
| 219 | + } |
| 220 | + results = pd.concat( |
| 221 | + [success_scaffold_count, pd.DataFrame(failed_scaffold_count)] |
| 222 | + ).sort_values("pdb_name") |
| 223 | + results.to_csv(os.path.join(output_dir, "result.csv")) |
| 224 | + print(results) |
| 225 | + |
| 226 | + |
| 227 | +def main(): |
| 228 | + parser = argparse.ArgumentParser() |
| 229 | + |
| 230 | + parser.add_argument("--scaffold_dir", type=str, default="./generation-results") |
| 231 | + |
| 232 | + args = parser.parse_args() |
| 233 | + |
| 234 | + motif_evaluation(args) |
| 235 | + |
| 236 | + |
| 237 | +if __name__ == "__main__": |
| 238 | + main() |
0 commit comments