COACH-D2.0 Evaluation Workflow
This file provides a step-by-step guide on how to run the COACH-D2.0 evaluation script to quantitatively assess predicted ligand-binding pockets for a given protein dataset.

Step 1: Prepare the Evaluation Directory and Extract Files
    # Create the evaluation working directory
    mkdir COACH-D2.0
    
    # Unzip the prediction results package (e.g., Coach420.zip) into the working directory
    unzip Coach420.zip -d COACH-D2.0

    After extraction, the directory structure should look like this:
    COACH-D2.0/
    └── Coach420/
        ├── label.csv              # Ground truth binding site annotations
        ├── protein.list           # List of protein IDs to be evaluated
        ├── prediction/
        │   └── COACH-D1.0_pred.txt
        │   └── COACH-D2.0_pred.txt
        │   └── DeepPocket_pred.txt
        │   └── GRaSP-web_pred.txt
        │   └── Prankweb4_pred.txt
        │   └── PUResNetV2.0_pred.txt
        ├── receptor_pdb/          # Receptor structure files in PDB format
        ├── ligand_pdb/            # Ligand structure files in PDB format
        ├── Evaluation_result.csv  # All method evaluations are summarized here
        └── PUResNetV2.0_Coach59_protein.list  # List of protein IDs to be evaluated for PUResNetV2.0 (excluding protein structures present in the PUResNetV2.0 training dataset)
        
        
Step 2: Run the Evaluation Script
    # Navigate to the extracted directory
    cd COACH-D2.0/Coach420
    
    # Run the COACH-D2.0 evaluation script
    python COACH-D2.0_evaluate.py --label_csv label.csv --protein_list protein.list --pred_txt prediction/COACH-D2.0_pred.txt --pdb_dir receptor_pdb --out_csv result/COACH-D2.0_result.csv --extra_pockets 0
      
      Parameter Descriptions:
            Argument	                                       Description
        --label_csv	      A CSV file containing ground truth annotations (e.g., residue-level binding sites)
        --protein_list	  A plain text file listing the protein IDs to be evaluated (one per line)
        --pred_txt	      Prediction file generated by the model, containing predicted pockets
        --pdb_dir	        Directory containing the receptor structures in PDB format
        --out_csv	        Path to the output CSV file that will store evaluation metrics
        --extra_pockets	  Optional. Specifies how many additional predicted pockets to consider beyond the number of true pockets (e.g., 2 means topN+2 pockets will be evaluated)
