/smina-molecular-docking

name: "smina-molecular-docking"
description: "smina molecular docking CLI. AutoDock Vina fork with customizable scoring functions, native SDF/MOL2/PDB ligand input, autoboxing, local energy minimization, and per-atom score breakdowns. Pipeline: receptor PDBQT prep -> ligand prep (RDKit/OpenBabel) -> dock via autobox or explicit grid -> rescore/minimize with custom scoring -> rank poses by affinity. Choose smina over Vina when you need custom scoring terms (--custom_scoring), local optimization of an existing pose (--local_only), per-atom contributions (--atom_term_data), or SDF/MOL2 ligands without manual PDBQT conversion. For unknown binding sites use diffdock; for the Python-bindings/Vinardo workflow use autodock-vina-docking."
license: "GPL-2.0"

smina Molecular Docking

Overview

smina is an AutoDock Vina 1.1.2 fork focused on flexible scoring and minimization. Accepts SDF/MOL2/PDB ligands directly (no manual PDBQT), autoboxes from a reference ligand, ships six built-in scoring functions plus arbitrary `--custom_scoring` terms, and prints per-atom score contributions. CLI-only — drive from Python via `subprocess`.

When to Use

• Re-scoring or locally minimizing an existing pose (`--local_only`, `--minimize`) without a full search
• Single-pose binding energy without docking (`--score_only`)
• Docking with a custom or empirical scoring function tuned to a target class
• SDF/MOL2/multi-ligand input without per-ligand PDBQT conversion
• Autoboxing the grid around a co-crystallized reference ligand
• Per-atom energy decomposition (`--atom_term_data`) for medchem analog design
• Batch virtual screening that parallelizes well across nodes (one CLI process per ligand)
• Use **autodock-vina-docking** instead when you need Vina Python bindings, Vinardo scoring, or Vina 1.2's expanded force field; use **diffdock** when the binding site is unknown

Prerequisites

• **smina binary** — conda-forge or SourceForge build
• **Python**: `rdkit`, `openbabel-wheel` (or system `openbabel`), `prody`, `pandas`, `py3Dmol`
• **ADFR Suite** for `prepare_receptor` (receptor PDBQT only — ligands handled by smina)
• **Data**: protein (PDB / PDB ID), ligand(s) as SMILES / SDF / MOL2

Check before installing — inside a pixi/conda env smina is usually already on PATH. If `command -v smina` succeeds, skip install; inside a pixi project invoke as `pixi run smina ...`.

command -v smina || conda install -c conda-forge smina openbabel
pip install rdkit prody pandas py3Dmol
# ADFR Suite: https://ccsb.scripps.edu/adfr/downloads/

Quick Start

End-to-end docking using autobox from a reference ligand:

import subprocess

result = subprocess.run([
    "smina",
    "-r", "1hpv_receptor.pdbqt",
    "-l", "candidate.sdf",
    "--autobox_ligand", "1hpv_ref_ligand.pdb",
    "--autobox_add", "8",          # padding around reference (Å)
    "-o", "candidate_docked.sdf",
    "--exhaustiveness", "16",
    "--num_modes", "9",
    "--seed", "42",
], check=True, capture_output=True, text=True)

print(result.stdout.splitlines()[-15:])  # affinity table at stdout tail

Workflow

Step 1: Prepare the Receptor (PDBQT)

Strip waters/hetatms, then run ADFR Suite's `prepare_receptor`.

import subprocess, prody

pdb_id = "1HPV"
prody.fetchPDB(pdb_id, compressed=False)
protein = prody.parsePDB(f"{pdb_id}.pdb").select("protein")
prody.writePDB(f"{pdb_id}_protein.pdb", protein)

receptor_pdbqt = f"{pdb_id}_receptor.pdbqt"
subprocess.run([
    "prepare_receptor",
    "-r", f"{pdb_id}_protein.pdb",
    "-o", receptor_pdbqt,
    "-A", "hydrogens",
], check=True)
print(f"Receptor: {receptor_pdbqt} ({protein.numAtoms()} atoms)")

Step 2: Prepare the Ligand (SDF)

smina reads SDF directly. Generate 3D coords with RDKit.

from rdkit import Chem
from rdkit.Chem import AllChem

mol = Chem.MolFromSmiles("CC(C)(C)NC(=O)[C@@H]1CN(CCc2ccccc2)C[C@H]1O")
mol = Chem.AddHs(mol)
AllChem.EmbedMolecule(mol, randomSeed=42)
AllChem.MMFFOptimizeMolecule(mol)

w = Chem.SDWriter("candidate.sdf"); w.write(mol); w.close()
print(f"Ligand SDF: candidate.sdf ({mol.GetNumAtoms()} atoms)")

Step 3: Extract a Reference Ligand for Autoboxing

`--autobox_ligand` derives the grid from a reference structure.

import prody

ref = prody.parsePDB(f"{pdb_id}.pdb").select("hetero and not water and not ion")
if ref is None:
    raise RuntimeError("No reference ligand — supply explicit --center_x/--size_x")
prody.writePDB(f"{pdb_id}_ref_ligand.pdb", ref)
print(f"Ref ligand: {ref.numAtoms()} atoms, center {ref.getCoords().mean(axis=0).round(2)}")

Step 4: Run Docking with Autobox

Affinity table is printed to stdout — capture it.

import subprocess

proc = subprocess.run([
    "smina",
    "-r", receptor_pdbqt,
    "-l", "candidate.sdf",
    "--autobox_ligand", f"{pdb_id}_ref_ligand.pdb",
    "--autobox_add", "8",
    "-o", "candidate_docked.sdf",
    "--exhaustiveness", "16",
    "--num_modes", "9",
    "--energy_range", "3",
    "--cpu", "4",
    "--seed", "42",
], check=True, capture_output=True, text=True)

for line in proc.stdout.splitlines()[-15:]:
    print(line)

Step 5: Parse Poses and Affinities

Affinities go into the SDF `<minimizedAffinity>` property.

from rdkit import Chem
import pandas as pd

rows = []
for i, mol in enumerate(Chem.SDMolSupplier("candidate_docked.sdf", removeHs=False)):
    if mol is None:
        continue
    aff = float(mol.GetProp("minimizedAffinity")) if mol.HasProp("minimizedAffinity") else None
    rmsd = float(mol.GetProp("minimizedRMSD")) if mol.HasProp("minimizedRMSD") else None
    rows.append({"pose": i + 1, "affinity_kcal_mol": aff, "rmsd_to_best": rmsd})

df = pd.DataFrame(rows).sort_values("affinity_kcal_mol")
print(df.to_string(index=False))
print(f"Best: {df.iloc[0]['affinity_kcal_mol']:.2f} kcal/mol")

Step 6: Local Minimization of an Existing Pose

`--local_only` refines an input pose without global