How to run inference with individual methods¶
How to run inference with DiffDock¶
Prepare CSV input files
python3 posebench/data/diffdock_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/diffdock_input_preparation.py dataset=astex_diverse
python3 posebench/data/diffdock_input_preparation.py dataset=dockgen
python3 posebench/data/diffdock_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets input_protein_structure_dir=data/casp15_set/casp15_holo_aligned_predicted_structures
Run inference on each dataset
python3 posebench/models/diffdock_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/diffdock_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/diffdock_inference.py dataset=dockgen repeat_index=1
...
python3 posebench/models/diffdock_inference.py dataset=casp15 batch_size=1 repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=diffdock dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=diffdock dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=diffdock dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=diffdock dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=diffdock dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=diffdock dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[diffdock\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_diffdock_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[diffdock\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_diffdock_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=diffdock dataset=casp15 repeat_index=1
...
How to run inference with FABind¶
Prepare CSV input files
python3 posebench/data/fabind_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/fabind_input_preparation.py dataset=astex_diverse
python3 posebench/data/fabind_input_preparation.py dataset=dockgen
Run inference on each dataset
python3 posebench/models/fabind_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/fabind_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/fabind_inference.py dataset=dockgen repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=fabind dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=fabind dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=fabind dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=fabind dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=fabind dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=fabind dataset=dockgen repeat_index=1
...
How to run inference with DynamicBind¶
Prepare CSV input files
python3 posebench/data/dynamicbind_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/dynamicbind_input_preparation.py dataset=astex_diverse
python3 posebench/data/dynamicbind_input_preparation.py dataset=dockgen
python3 posebench/data/dynamicbind_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets
Run inference on each dataset
python3 posebench/models/dynamicbind_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/dynamicbind_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/dynamicbind_inference.py dataset=dockgen repeat_index=1
...
python3 posebench/models/dynamicbind_inference.py dataset=casp15 batch_size=1 input_data_dir=data/casp15_set/casp15_holo_aligned_predicted_structures repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=dynamicbind dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=dynamicbind dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=dynamicbind dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=dynamicbind dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=dynamicbind dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=dynamicbind dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[dynamicbind\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_dynamicbind_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[dynamicbind\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_dynamicbind_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=dynamicbind dataset=casp15 repeat_index=1
...
How to run inference with NeuralPLexer¶
Prepare CSV input files
python3 posebench/data/neuralplexer_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/neuralplexer_input_preparation.py dataset=astex_diverse
python3 posebench/data/neuralplexer_input_preparation.py dataset=dockgen
python3 posebench/data/neuralplexer_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets input_receptor_structure_dir=data/casp15_set/casp15_holo_aligned_predicted_structures
Run inference on each dataset
python3 posebench/models/neuralplexer_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/neuralplexer_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/neuralplexer_inference.py dataset=dockgen repeat_index=1
...
python3 posebench/models/neuralplexer_inference.py dataset=casp15 chunk_size=5 repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=neuralplexer dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=neuralplexer dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=neuralplexer dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Align predicted protein-ligand structures to ground-truth complex structures
conda activate PyMOL-PoseBench
python3 posebench/analysis/complex_alignment.py method=neuralplexer dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=neuralplexer dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=neuralplexer dataset=dockgen repeat_index=1
...
conda deactivate
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=neuralplexer dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=neuralplexer dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=neuralplexer dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[neuralplexer\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_neuralplexer_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[neuralplexer\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_neuralplexer_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=neuralplexer dataset=casp15 repeat_index=1
...
How to run inference with FlowDock¶
Prepare CSV input files
python3 posebench/data/flowdock_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/flowdock_input_preparation.py dataset=astex_diverse
python3 posebench/data/flowdock_input_preparation.py dataset=dockgen
python3 posebench/data/flowdock_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets input_receptor_structure_dir=data/casp15_set/casp15_holo_aligned_predicted_structures
Run inference on each dataset
python3 posebench/models/flowdock_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/flowdock_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/flowdock_inference.py dataset=dockgen repeat_index=1
...
python3 posebench/models/flowdock_inference.py dataset=casp15 chunk_size=5 repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=flowdock dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=flowdock dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=flowdock dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Align predicted protein-ligand structures to ground-truth complex structures
conda activate PyMOL-PoseBench
python3 posebench/analysis/complex_alignment.py method=flowdock dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=flowdock dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=flowdock dataset=dockgen repeat_index=1
...
conda deactivate
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=flowdock dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=flowdock dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=flowdock dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[flowdock\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_flowdock_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[flowdock\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_flowdock_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=flowdock dataset=casp15 repeat_index=1
...
How to run inference with RoseTTAFold-All-Atom¶
Prepare CSV input files
python3 posebench/data/rfaa_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/rfaa_input_preparation.py dataset=astex_diverse
python3 posebench/data/rfaa_input_preparation.py dataset=dockgen
python3 posebench/data/rfaa_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets
Run inference on each dataset
conda activate forks/RoseTTAFold-All-Atom/RFAA/
python3 posebench/models/rfaa_inference.py dataset=posebusters_benchmark run_inference_directly=true
python3 posebench/models/rfaa_inference.py dataset=astex_diverse run_inference_directly=true
python3 posebench/models/rfaa_inference.py dataset=dockgen run_inference_directly=true
python3 posebench/models/rfaa_inference.py dataset=casp15 run_inference_directly=true
conda deactivate
Extract predictions into separate files for proteins and ligands
python3 posebench/data/rfaa_output_extraction.py dataset=posebusters_benchmark
python3 posebench/data/rfaa_output_extraction.py dataset=astex_diverse
python3 posebench/data/rfaa_output_extraction.py dataset=dockgen
python3 posebench/data/rfaa_output_extraction.py dataset=casp15
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=rfaa dataset=posebusters_benchmark remove_initial_protein_hydrogens=true
python3 posebench/models/inference_relaxation.py method=rfaa dataset=astex_diverse remove_initial_protein_hydrogens=true
python3 posebench/models/inference_relaxation.py method=rfaa dataset=dockgen remove_initial_protein_hydrogens=true
Align predicted protein-ligand structures to ground-truth complex structures
conda activate PyMOL-PoseBench
python3 posebench/analysis/complex_alignment.py method=rfaa dataset=posebusters_benchmark
python3 posebench/analysis/complex_alignment.py method=rfaa dataset=astex_diverse
python3 posebench/analysis/complex_alignment.py method=rfaa dataset=dockgen
conda deactivate
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=rfaa dataset=posebusters_benchmark
python3 posebench/analysis/inference_analysis.py method=rfaa dataset=astex_diverse
python3 posebench/analysis/inference_analysis.py method=rfaa dataset=dockgen
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[rfaa\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_rfaa_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[rfaa\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_rfaa_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=rfaa dataset=casp15 repeat_index=1
...
How to run inference with Chai-1¶
Prepare CSV input files
python3 posebench/data/chai_input_preparation.py dataset=posebusters_benchmark
python3 posebench/data/chai_input_preparation.py dataset=astex_diverse
python3 posebench/data/chai_input_preparation.py dataset=dockgen
python3 posebench/data/chai_input_preparation.py dataset=casp15 input_data_dir=data/casp15_set/targets
Run inference on each dataset
conda activate forks/chai-lab/chai-lab/
python3 posebench/models/chai_inference.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/models/chai_inference.py dataset=astex_diverse repeat_index=1
...
python3 posebench/models/chai_inference.py dataset=dockgen repeat_index=1
...
python3 posebench/models/chai_inference.py dataset=casp15 repeat_index=1
...
conda deactivate
Extract predictions into separate files for proteins and ligands
python3 posebench/data/chai_output_extraction.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/data/chai_output_extraction.py dataset=astex_diverse repeat_index=1
...
python3 posebench/data/chai_output_extraction.py dataset=dockgen repeat_index=1
...
python3 posebench/data/chai_output_extraction.py dataset=casp15 repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=chai-lab dataset=posebusters_benchmark remove_initial_protein_hydrogens=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=chai-lab dataset=astex_diverse remove_initial_protein_hydrogens=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=chai-lab dataset=dockgen remove_initial_protein_hydrogens=true repeat_index=1
...
Align predicted protein-ligand structures to ground-truth complex structures
conda activate PyMOL-PoseBench
python3 posebench/analysis/complex_alignment.py method=chai-lab dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=chai-lab dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=chai-lab dataset=dockgen repeat_index=1
conda deactivate
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=chai-lab dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=chai-lab dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=chai-lab dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[chai-lab\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_chai-lab_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[chai-lab\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_chai-lab_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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
...
How to run inference with AlphaFold 3¶
Run inference (3x) using the academically-available inference code released on GitHub, saving each run’s structures to a unique output directory located at forks/alphafold3/prediction_outputs/{dataset=posebusters_benchmark,astex_diverse,dockgen,casp15}_{repeat_index=1,2,3}
Then, extract predictions into separate files for proteins and ligands
python3 posebench/data/af3_output_extraction.py dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/data/af3_output_extraction.py dataset=astex_diverse repeat_index=1
...
python3 posebench/data/af3_output_extraction.py dataset=dockgen repeat_index=1
...
python3 posebench/data/af3_output_extraction.py dataset=casp15 repeat_index=1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=alphafold3 dataset=posebusters_benchmark remove_initial_protein_hydrogens=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=alphafold3 dataset=astex_diverse remove_initial_protein_hydrogens=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=alphafold3 dataset=dockgen remove_initial_protein_hydrogens=true repeat_index=1
...
Align predicted protein-ligand structures to ground-truth complex structures
conda activate PyMOL-PoseBench
python3 posebench/analysis/complex_alignment.py method=alphafold3 dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=alphafold3 dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/complex_alignment.py method=alphafold3 dataset=dockgen repeat_index=1
conda deactivate
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=alphafold3 dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=alphafold3 dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=alphafold3 dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# first assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[alphafold3\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_alphafold3_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[alphafold3\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_alphafold3_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=alphafold3 dataset=casp15 repeat_index=1
...
How to run inference with AutoDock Vina¶
Prepare CSV input files
cp forks/DiffDock/inference/diffdock_posebusters_benchmark_inputs.csv forks/Vina/inference/vina_posebusters_benchmark_inputs.csv
cp forks/DiffDock/inference/diffdock_astex_diverse_inputs.csv forks/Vina/inference/vina_astex_diverse_inputs.csv
cp forks/DiffDock/inference/diffdock_dockgen_inputs.csv forks/Vina/inference/vina_dockgen_inputs.csv
cp forks/DiffDock/inference/diffdock_casp15_inputs.csv forks/Vina/inference/vina_casp15_inputs.csv
Run inference on each dataset
python3 posebench/models/vina_inference.py dataset=posebusters_benchmark method=p2rank repeat_index=1 # NOTE: P2Rank's binding pockets are recommended as the default Vina input
...
python3 posebench/models/vina_inference.py dataset=astex_diverse method=p2rank repeat_index=1
...
python3 posebench/models/vina_inference.py dataset=dockgen method=p2rank repeat_index=1
...
python3 posebench/models/vina_inference.py dataset=casp15 method=p2rank repeat_index=1
...
Copy Vina’s predictions to the corresponding inference directory for each repeat
mkdir -p forks/Vina/inference/vina_p2rank_posebusters_benchmark_outputs_1 && cp -r data/test_cases/posebusters_benchmark/vina_p2rank_posebusters_benchmark_outputs_1/* forks/Vina/inference/vina_p2rank_posebusters_benchmark_outputs_1
...
mkdir -p forks/Vina/inference/vina_p2rank_astex_diverse_outputs_1 && cp -r data/test_cases/astex_diverse/vina_p2rank_astex_diverse_outputs_1/* forks/Vina/inference/vina_p2rank_astex_diverse_outputs_1
...
mkdir -p forks/Vina/inference/vina_p2rank_dockgen_outputs_1 && cp -r data/test_cases/dockgen/vina_p2rank_dockgen_outputs_1/* forks/Vina/inference/vina_p2rank_dockgen_outputs_1
...
mkdir -p forks/Vina/inference/vina_p2rank_casp15_outputs_1 && cp -r data/test_cases/casp15/vina_p2rank_casp15_outputs_1/* forks/Vina/inference/vina_p2rank_casp15_outputs_1
...
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=vina vina_binding_site_method=p2rank dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=vina vina_binding_site_method=p2rank dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
python3 posebench/models/inference_relaxation.py method=vina vina_binding_site_method=p2rank dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true repeat_index=1
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=vina vina_binding_site_method=p2rank dataset=posebusters_benchmark repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=vina vina_binding_site_method=p2rank dataset=astex_diverse repeat_index=1
...
python3 posebench/analysis/inference_analysis.py method=vina vina_binding_site_method=p2rank dataset=dockgen repeat_index=1
...
Analyze inference results for the CASP15 dataset
# assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[vina\] vina_binding_site_methods=\[p2rank\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_vina_p2rank_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[vina\] vina_binding_site_methods=\[p2rank\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_vina_p2rank_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=vina vina_binding_site_method=p2rank dataset=casp15 repeat_index=1
...
How to run inference with TULIP¶
Gather all template ligands generated by TULIP via its dedicated GitHub repository and collate the resulting ligand fragment SDF files
python3 posebench/data/tulip_output_extraction.py dataset=posebusters_benchmark
python3 posebench/data/tulip_output_extraction.py dataset=astex_diverse
python3 posebench/data/tulip_output_extraction.py dataset=dockgen
python3 posebench/data/tulip_output_extraction.py dataset=casp15
Relax the generated ligand structures inside of their respective protein pockets
python3 posebench/models/inference_relaxation.py method=tulip dataset=posebusters_benchmark remove_initial_protein_hydrogens=true assign_partial_charges_manually=true
...
python3 posebench/models/inference_relaxation.py method=tulip dataset=astex_diverse remove_initial_protein_hydrogens=true assign_partial_charges_manually=true
...
python3 posebench/models/inference_relaxation.py method=tulip dataset=dockgen remove_initial_protein_hydrogens=true assign_partial_charges_manually=true
...
Analyze inference results for each dataset
python3 posebench/analysis/inference_analysis.py method=tulip dataset=posebusters_benchmark
...
python3 posebench/analysis/inference_analysis.py method=tulip dataset=astex_diverse
...
python3 posebench/analysis/inference_analysis.py method=tulip dataset=dockgen
...
Analyze inference results for the CASP15 dataset
# then assemble (unrelaxed and post ranking-relaxed) CASP15-compliant prediction submission files for scoring
python3 posebench/models/ensemble_generation.py ensemble_methods=\[tulip\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_tulip_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=false export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
python3 posebench/models/ensemble_generation.py ensemble_methods=\[tulip\] input_csv_filepath=data/test_cases/casp15/ensemble_inputs.csv output_dir=data/test_cases/casp15/top_tulip_ensemble_predictions_1 skip_existing=true relax_method_ligands_post_ranking=true export_file_format=casp15 export_top_n=5 combine_casp_output_files=true max_method_predictions=5 method_top_n_to_select=5 resume=true ensemble_benchmarking=true ensemble_benchmarking_dataset=casp15 cuda_device_index=0 ensemble_benchmarking_repeat_index=1
# 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=tulip dataset=casp15
...