bio-virtual-screening

Ligand Preparation

Goal: Convert a SMILES string into a docking-ready 3D ligand file.

Approach: Generate a 3D conformer with ETKDGv3, optimize geometry with MMFF, write to MOL, and convert to PDBQT with Gasteiger charges via Open Babel.

Docking with Vina

Goal: Dock a single ligand into a protein binding site and retrieve predicted binding affinities.

Approach: Initialize Vina with the receptor, set the search space around the binding site, dock with specified exhaustiveness, and extract ranked poses with energies.

from vina import Vina

def dock_ligand(receptor_pdbqt, ligand_pdbqt, center, box_size, exhaustiveness=8):
    '''
    Dock a single ligand using AutoDock Vina 1.2.

    Args:
        receptor_pdbqt: Prepared receptor file
        ligand_pdbqt: Prepared ligand file
        center: (x, y, z) center of binding site
        box_size: (x, y, z) box dimensions (Angstroms)
        exhaustiveness: Search thoroughness (8=quick, 32=production, 64=thorough)
    '''
    v = Vina(sf_name='vina')
    v.set_receptor(receptor_pdbqt)
    v.set_ligand_from_file(ligand_pdbqt)

    # Define search space
    # Box size generally < 30x30x30 Angstroms
    v.compute_vina_maps(center=center, box_size=box_size)

    # Dock
    v.dock(exhaustiveness=exhaustiveness, n_poses=10)

    # Get results
    energies = v.energies()  # List of (affinity, rmsd_lb, rmsd_ub)
    poses = v.poses()  # PDBQT string of all poses

    return energies, poses

## Virtual Screening Pipeline

**Goal:** Screen an entire compound library against a protein target and rank by binding affinity.

**Approach:** Prepare each ligand from SMILES, dock against the pre-computed receptor maps, save top poses, and compile results into a sorted DataFrame.