scripts/extract_features.py

#!/usr/bin/env python3
"""
Extract structural and sequence features from AlphaFold PDB files.
Part of APED - African Protein Engineering Dataset
"""

import argparse
import os
import warnings
from pathlib import Path

import numpy as np
import pandas as pd

warnings.filterwarnings('ignore')

# Amino acid properties
HYDROPHOBICITY = {
    'A': 1.8, 'R': -4.5, 'N': -3.5, 'D': -3.5, 'C': 2.5,
    'Q': -3.5, 'E': -3.5, 'G': -0.4, 'H': -3.2, 'I': 4.5,
    'L': 3.8, 'K': -3.9, 'M': 1.9, 'F': 2.8, 'P': -1.6,
    'S': -0.8, 'T': -0.7, 'W': -0.9, 'Y': -1.3, 'V': 4.2
}

CHARGE = {
    'A': 0, 'R': 1, 'N': 0, 'D': -1, 'C': 0,
    'Q': 0, 'E': -1, 'G': 0, 'H': 0.5, 'I': 0,
    'L': 0, 'K': 1, 'M': 0, 'F': 0, 'P': 0,
    'S': 0, 'T': 0, 'W': 0, 'Y': 0, 'V': 0
}

MW = {
    'A': 89.1, 'R': 174.2, 'N': 132.1, 'D': 133.1, 'C': 121.2,
    'Q': 146.2, 'E': 147.1, 'G': 75.1, 'H': 155.2, 'I': 131.2,
    'L': 131.2, 'K': 146.2, 'M': 149.2, 'F': 165.2, 'P': 115.1,
    'S': 105.1, 'T': 119.1, 'W': 204.2, 'Y': 181.2, 'V': 117.1
}


def parse_pdb(pdb_file: Path) -> dict:
    """Parse PDB file and extract sequence, pLDDT, and secondary structure."""
    sequence = []
    plddt_scores = []
    residue_ids = set()
    
    three_to_one = {
        'ALA': 'A', 'ARG': 'R', 'ASN': 'N', 'ASP': 'D', 'CYS': 'C',
        'GLN': 'Q', 'GLU': 'E', 'GLY': 'G', 'HIS': 'H', 'ILE': 'I',
        'LEU': 'L', 'LYS': 'K', 'MET': 'M', 'PHE': 'F', 'PRO': 'P',
        'SER': 'S', 'THR': 'T', 'TRP': 'W', 'TYR': 'Y', 'VAL': 'V'
    }
    
    with open(pdb_file, 'r') as f:
        for line in f:
            if line.startswith('ATOM') and line[12:16].strip() == 'CA':
                res_id = int(line[22:26].strip())
                if res_id not in residue_ids:
                    residue_ids.add(res_id)
                    res_name = line[17:20].strip()
                    if res_name in three_to_one:
                        sequence.append(three_to_one[res_name])
                        plddt = float(line[60:66].strip())
                        plddt_scores.append(plddt)
    
    return {
        'sequence': ''.join(sequence),
        'plddt_scores': plddt_scores
    }


def calculate_sequence_features(sequence: str) -> dict:
    """Calculate sequence-based features."""
    length = len(sequence)
    if length == 0:
        return {}
    
    # Hydrophobicity
    hydro = np.mean([HYDROPHOBICITY.get(aa, 0) for aa in sequence])
    
    # Net charge
    charge = sum([CHARGE.get(aa, 0) for aa in sequence])
    
    # Molecular weight
    mw = sum([MW.get(aa, 0) for aa in sequence]) - 18.015 * (length - 1)
    
    # Isoelectric point (simplified)
    pos_count = sum(1 for aa in sequence if aa in 'RKH')
    neg_count = sum(1 for aa in sequence if aa in 'DE')
    pi = 7.0 + (pos_count - neg_count) * 0.5 / max(length, 1)
    pi = max(3.0, min(12.0, pi))
    
    # Instability index (simplified)
    dipeptide_instability = 0
    for i in range(len(sequence) - 1):
        if sequence[i:i+2] in ['DG', 'GD', 'NG', 'GN']:
            dipeptide_instability += 1
    instability = 40 + dipeptide_instability * 10 / max(length, 1)
    
    # Aromaticity
    aromatic = sum(1 for aa in sequence if aa in 'FWY') / length
    
    # Amino acid composition
    aa_counts = {aa: sequence.count(aa) / length for aa in 'ACDEFGHIKLMNPQRSTVWY'}
    
    return {
        'sequence_length': length,
        'hydrophobicity': round(hydro, 4),
        'charge': charge,
        'molecular_weight': round(mw, 2),
        'isoelectric_point': round(pi, 2),
        'instability_index': round(instability, 2),
        'aromaticity': round(aromatic, 4),
        'proline_fraction': round(aa_counts.get('P', 0), 4),
        'glycine_fraction': round(aa_counts.get('G', 0), 4),
        'charged_fraction': round(sum(aa_counts.get(aa, 0) for aa in 'RKDE'), 4)
    }


def estimate_secondary_structure(sequence: str) -> dict:
    """Estimate secondary structure propensities."""
    helix_formers = set('AELM')
    sheet_formers = set('VIY')
    
    helix_count = sum(1 for aa in sequence if aa in helix_formers)
    sheet_count = sum(1 for aa in sequence if aa in sheet_formers)
    length = max(len(sequence), 1)
    
    helix_frac = helix_count / length
    sheet_frac = sheet_count / length
    coil_frac = 1.0 - helix_frac - sheet_frac
    
    return {
        'helix_fraction': round(max(0, helix_frac), 4),
        'sheet_fraction': round(max(0, sheet_frac), 4),
        'coil_fraction': round(max(0, coil_frac), 4)
    }


def extract_features_from_pdb(pdb_file: Path) -> dict:
    """Extract all features from a single PDB file."""
    # Get UniProt ID from filename
    filename = pdb_file.stem
    if filename.startswith('AF-'):
        uniprot_id = filename.split('-')[1]
    else:
        uniprot_id = filename
    
    # Parse PDB
    pdb_data = parse_pdb(pdb_file)
    sequence = pdb_data['sequence']
    plddt_scores = pdb_data['plddt_scores']
    
    if not sequence:
        return None
    
    # Calculate features
    features = {
        'uniprot_id': uniprot_id,
        'sequence': sequence,
        'mean_plddt': round(np.mean(plddt_scores), 2) if plddt_scores else 0,
        'min_plddt': round(np.min(plddt_scores), 2) if plddt_scores else 0,
        'max_plddt': round(np.max(plddt_scores), 2) if plddt_scores else 0,
    }
    
    # Add sequence features
    features.update(calculate_sequence_features(sequence))
    
    # Add secondary structure estimates
    features.update(estimate_secondary_structure(sequence))
    
    return features


def main():
    parser = argparse.ArgumentParser(
        description="Extract features from AlphaFold PDB structures"
    )
    parser.add_argument(
        "--input-dir",
        type=str,
        required=True,
        help="Directory containing PDB files"
    )
    parser.add_argument(
        "--output",
        type=str,
        default="data/ml_ready/aped_ml_dataset.parquet",
        help="Output file path (parquet or csv)"
    )
    parser.add_argument(
        "--min-plddt",
        type=float,
        default=50.0,
        help="Minimum mean pLDDT to include"
    )
    
    args = parser.parse_args()
    
    input_dir = Path(args.input_dir)
    output_path = Path(args.output)
    output_path.parent.mkdir(parents=True, exist_ok=True)
    
    pdb_files = list(input_dir.glob("*.pdb"))
    print(f"Found {len(pdb_files)} PDB files")
    
    features_list = []
    
    for i, pdb_file in enumerate(pdb_files):
        features = extract_features_from_pdb(pdb_file)
        
        if features and features['mean_plddt'] >= args.min_plddt:
            features_list.append(features)
        
        if (i + 1) % 100 == 0:
            print(f"  Processed {i + 1}/{len(pdb_files)}")
    
    df = pd.DataFrame(features_list)
    
    # Save
    if str(output_path).endswith('.parquet'):
        df.to_parquet(output_path, index=False)
    else:
        df.to_csv(output_path, index=False)
    
    print(f"\nComplete!")
    print(f"  Proteins: {len(df)}")
    print(f"  Features: {len(df.columns)}")
    print(f"  Output: {output_path}")
    
    # Print summary statistics
    print(f"\nFeature summary:")
    print(f"  Mean pLDDT: {df['mean_plddt'].mean():.1f} ± {df['mean_plddt'].std():.1f}")
    print(f"  Mean length: {df['sequence_length'].mean():.0f} ± {df['sequence_length'].std():.0f}")


if __name__ == "__main__":
    main()