Fix complex alignment for Chai-1

BioinfoMachineLearning · Sep 26, 2024 · f598865 · f598865
1 parent ac8b5c3
commit f598865
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 5 deletions.
diff --git a/README.md b/README.md
@@ -740,7 +740,7 @@ python3 posebench/models/ensemble_generation.py ensemble_methods=\[chai-lab\] in
 # NOTE: the suffixes for both `output_dir` and `ensemble_benchmarking_repeat_index` should be modified to e.g., 2, 3, ...
 ...
 # now score the CASP15-compliant submissions using the official CASP scoring pipeline
-python3 posebench/analysis/inference_analysis_casp.py method=chai-lab dataset=casp15 repeat_index=1
+python3 posebench/analysis/inference_analysis_casp.py method=chai-lab dataset=casp15 repeat_index=1 targets='[H1135, T1127v2, T1146, T1152, T1158v1, T1158v2, T1158v3, T1158v4, T1186, T1187, T1188]'
 ...
 ```
 

diff --git a/posebench/analysis/complex_alignment.py b/posebench/analysis/complex_alignment.py
@@ -411,8 +411,11 @@ def main(cfg: DictConfig):
                 output_protein_files = sorted(
                     [
                         item
-                        for item in output_dir.rglob("*protein.pdb")
-                        if any(
+                        for item in output_dir.rglob(
+                            f"pred.model_idx_{cfg.rank_to_align - 1}_protein*.pdb"
+                        )
+                        if "_aligned" not in item.stem
+                        and any(
                             [item.parent.stem in file.parent.stem for file in output_ligand_files]
                         )
                     ]
@@ -426,7 +429,9 @@ def main(cfg: DictConfig):
         ), f"Numbers of protein ({len(output_protein_files)}) and ligand ({len(output_ligand_files)}) files do not match."
 
         # align protein-ligand complexes
-        for protein_file, ligand_file in tqdm(zip(output_protein_files, output_ligand_files)):
+        for protein_file, ligand_file in tqdm(
+            zip(output_protein_files, output_ligand_files), desc="Aligning complexes"
+        ):
             protein_id, ligand_id = protein_file.stem, ligand_file.stem
             if protein_id != ligand_id and cfg.method == "dynamicbind":
                 protein_id, ligand_id = (

diff --git a/scripts/build_inference_script.py b/scripts/build_inference_script.py
@@ -156,7 +156,7 @@
             "python3 posebench/models/ensemble_generation.py ensemble_methods=[chai-lab] ensemble_ranking_method={ensemble_ranking_method} input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_chai-lab_ensemble_predictions_{repeat_index} skip_existing=true relax_method_ligands_post_ranking=true relax_protein={relax_protein} export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=40 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index={cuda_device_index} ensemble_benchmarking_repeat_index={repeat_index}",
         ],
         "analyze_casp15": [
-            "python3 posebench/analysis/inference_analysis_casp.py method=chai-lab dataset=casp15 repeat_index={repeat_index} relax_protein={relax_protein}",
+            "python3 posebench/analysis/inference_analysis_casp.py method=chai-lab dataset=casp15 repeat_index={repeat_index} relax_protein={relax_protein} targets='[H1135, T1127v2, T1146, T1152, T1158v1, T1158v2, T1158v3, T1158v4, T1186, T1187, T1188]'",
         ],
     },
     "vina": {