SMINA

Enhanced molecular docking with custom scoring functions and fast minimization

Docs

Input

Protein (Receptor)

Ligand

Scoring function

Docking mode

Exhaustiveness

Number of poses

Energy range (kcal/mol)

RMSD filter (Å)

Configure search space

Per-atom interaction energies

25 credits

Output

Configure input settings on the left, then click "Submit job"orLoad an example (it's free)

Quick docking test

Fast pose refinement

High-accuracy with Vinardo

What is SMINA?

SMINA is a fork of AutoDock Vina developed by David Koes at the University of Pittsburgh. It extends Vina with comprehensive support for custom scoring functions, dramatically faster minimization, and enhanced workflows for scoring function research.

Published in 2013, SMINA was designed to address limitations in Vina's scoring function development capabilities. The tool enables researchers to define custom scoring functions with user-specified weights, making it valuable for both standard docking and advanced scoring research.

SMINA also includes the Vinardo scoring function, an optimized variant that consistently outperforms Vina's default scoring in pose prediction, ranking, and virtual screening benchmarks.

How does SMINA work?

SMINA inherits Vina's core search algorithm while extending its scoring capabilities. The search uses Iterated Local Search with BFGS quasi-Newton optimization, exploring the conformational space through random mutations of position, orientation, and torsion angles.

Scoring functions

SMINA provides three scoring options:

Default (Vina) uses the original AutoDock Vina scoring function with Gaussian steric terms, hydrogen bonding, hydrophobic interactions, and a rotatable bond penalty. This is well-validated and suitable for most docking applications.

Vinardo removes Vina's problematic second Gaussian term that contributed 58% of binding energy despite creating artifacts. Vinardo uses optimized atomic radii and simplified terms, achieving 89.7% top-1 docking success versus Vina's 80.5% on the CASF-2013 benchmark.

Custom scoring allows researchers to define their own scoring functions using 58 available interaction terms, including electrostatics, desolvation, Lennard-Jones 4-8 potentials, and simple property counts.

Available scoring terms

SMINA's custom scoring supports these term types:

Steric: Gaussian functions (gauss), repulsion potentials
Chemical: Hydrogen bonding, hydrophobic contacts, non-hydrophobic contacts
Electrostatic: Distance-dependent electrostatic terms
Desolvation: AutoDock 4-style solvation model
Van der Waals: Lennard-Jones 4-8 formulation
Property counts: Torsion counts, hydrophobic atom counts

Each term accepts parameters for offset, width, cutoff, and exponent, enabling fine-grained control over the scoring function behavior.

Docking modes

SMINA provides four operational modes:

Mode	Description	Speed
`Full docking`	Global search + local optimization	Standard
`Minimize only`	Refine existing pose without global search	10-20x faster
`Score only`	Evaluate pose without moving atoms	Fastest
`Local only`	Local optimization from starting pose	Fast

The Minimize only mode is particularly valuable for pose refinement workflows, achieving 10-20x speedup over full docking by skipping global search.

Input requirements

Protein (Receptor)

Upload a PDB file or enter an RCSB PDB ID. The protein should be prepared with hydrogens added and missing residues fixed. Use PDB Fixer for automated preparation.

Ligand

Enter a SMILES string for small molecules, or upload an SDF, MOL, or PDBQT file. SMINA automatically generates 3D coordinates and calculates partial charges via OpenBabel.

For complex ligands (macrocycles, peptides, >150 atoms), pre-prepared PDBQT files are recommended to ensure proper handling.

Docking parameters

Scoring function

Default (Vina): Standard Vina scoring, well-validated for general docking
Vinardo: Optimized for pose prediction accuracy, recommended for structure-based drug design
Custom scoring file: Define your own scoring weights for research applications

Docking mode

Full docking: Complete global search with local optimization
Minimize only: Fast pose refinement (10-20x faster than full docking)
Score only: Evaluate a pose without any optimization
Local only: Optimize from the input starting pose

Search parameters

Exhaustiveness: Number of independent docking runs. Higher values (16-32) improve search coverage for difficult targets
Number of poses: Maximum binding poses to return. SMINA supports up to 50 poses, versus Vina's limit of 20
Energy range: Maximum energy difference from the best pose for inclusion. Only poses within this range are returned
RMSD filter: Minimum RMSD between poses to reduce redundancy

Search space

The search box defines where SMINA explores ligand binding:

Auto: Automatically calculates a box encompassing the entire protein surface
Autobox: Derives the search box from a reference ligand (useful when binding site is known from a co-crystal structure)
Manual: Specify exact center coordinates and box dimensions

We recommend using autobox when a reference ligand structure is available, as this focuses the search and improves both speed and accuracy.

Output options

Per-atom interaction energies: Outputs energy contributions for each atom, useful for analyzing which parts of the ligand contribute most to binding

Understanding the results

Binding affinity

Affinity is reported in kcal/mol. More negative values indicate stronger predicted binding:

Range	Interpretation
-4 to -6	Weak binding
-6 to -8	Moderate binding
-8 to -10	Strong binding
< -10	Very strong binding

Vinardo scores may differ slightly from Vina due to the modified scoring function. Both use the same scale and interpretation guidelines.

Pose ranking

SMINA ranks poses by predicted binding affinity. The top-ranked pose represents the most favorable binding mode, but examining multiple poses is recommended. Alternative binding modes within 1-2 kcal/mol of the best score may be biologically relevant.

Validation metrics

On the CASF-2013 benchmark, Vinardo achieves:

Docking power: 89.7% success rate (pose RMSD < 2Å from crystal structure)
Screening power: Identifies true binders in top 1% for 80% of proteins
Scoring correlation: r = 0.601 between predicted and experimental binding affinities

Comparison to other docking tools

Feature	SMINA	AutoDock Vina	GNINA
Scoring	Vina/Vinardo/Custom	Vina	Vina + CNN
Max poses	50	20	20
Minimization speed	10-20x faster	Standard	Standard
Custom scoring	Yes	No	No
Per-atom energies	Yes	No	No

SMINA is ideal when you need custom scoring functions, fast minimization, or more than 20 poses. For standard docking with known scoring, Vina or GNINA work equally well.

Custom scoring function syntax

Custom scoring functions are defined as weighted sums of interaction terms. Each line specifies a weight and term with parameters:

Text

-0.035579 gauss(o=0,_w=0.5,_c=8)
-0.005156 gauss(o=3,_w=2,_c=8)
0.840245 repulsion(o=0,_c=8)
-0.035069 hydrophobic(g=0.5,_b=1.5,_c=8)
-0.587439 non_dir_h_bond(g=-0.7,_b=0,_c=8)
1.923 num_tors_div

Parameters include:

o: Offset from optimal distance
w: Gaussian width
c: Distance cutoff
g: Good distance threshold
b: Bad distance threshold

Detailed term documentation is available in the SMINA GitHub repository.

Best practices

Use Vinardo for pose prediction: Vinardo consistently outperforms default Vina scoring in docking benchmarks
Define search space when possible: Using autobox with a reference ligand improves both speed and accuracy
Use minimize mode for refinement: When you have a reasonable starting pose, minimize mode is 10-20x faster than full docking
Increase exhaustiveness for difficult targets: Flexible ligands or large binding sites benefit from exhaustiveness values of 16-32

Related tools

PandaDock

Open-source molecular docking platform using physics-based scoring functions. CPU-optimized algorithms achieve sub-angstrom accuracy (0.014A RMSD) without GPU requirements.

AutoDock-GPU

GPU-accelerated molecular docking using the AutoDock4 force field. Up to 56x faster than serial AutoDock via CUDA parallelization of the Lamarckian Genetic Algorithm.

AutoDock Vina

AutoDock Vina is a widely-used molecular docking tool that predicts protein-ligand binding modes using physics-based force fields. Fast, reliable, and the gold standard for structure-based drug discovery.

FlowDock

FlowDock predicts protein-ligand complex structures and binding-affinity scores using geometric flow matching.

GNINA

GNINA is a molecular docking tool that combines traditional physics-based docking with deep learning CNN scoring for protein-small-molecule complexes. It provides accurate binding predictions with confidence scores, optimized for high-throughput virtual screening.

DiffDock-L

DiffDock-L is a state-of-the-art molecular docking tool that uses diffusion models to predict how small molecule ligands bind to protein targets. It generates multiple binding poses with confidence scores.

DynamicBind

DynamicBind is an AI-powered protein-ligand binding prediction tool that recovers ligand-induced conformational changes from unbound protein structures. It predicts both ligand binding poses and protein conformational changes.

SigmaDock

SigmaDock is a fragment-based molecular docking tool using SE(3) equivariant diffusion models to predict how small molecule ligands bind to protein targets. Presented at ICLR 2026, it generates multiple binding poses with Vinardo scoring.

SurfDock

SurfDock is a surface-informed diffusion generative model for protein-ligand docking, published in Nature Methods 2024. It leverages protein surface geometry to guide a diffusion process for reliable and accurate protein-ligand complex prediction.

OpenFE

Run alchemical free energy calculations for drug discovery using Open Free Energy. Supports Absolute Hydration Free Energy (AHFE) and Relative Binding Free Energy (RBFE) calculations with GPU-accelerated OpenMM simulations.

What is SMINA?

SMINA also includes the Vinardo scoring function, an optimized variant that consistently outperforms Vina's default scoring in pose prediction, ranking, and virtual screening benchmarks.

How does SMINA work?

Scoring functions

SMINA provides three scoring options:

Available scoring terms

SMINA's custom scoring supports these term types:

Steric: Gaussian functions (gauss), repulsion potentials
Chemical: Hydrogen bonding, hydrophobic contacts, non-hydrophobic contacts
Electrostatic: Distance-dependent electrostatic terms
Desolvation: AutoDock 4-style solvation model
Van der Waals: Lennard-Jones 4-8 formulation
Property counts: Torsion counts, hydrophobic atom counts

Each term accepts parameters for offset, width, cutoff, and exponent, enabling fine-grained control over the scoring function behavior.

Docking modes

SMINA provides four operational modes:

Mode	Description	Speed
`Full docking`	Global search + local optimization	Standard
`Minimize only`	Refine existing pose without global search	10-20x faster
`Score only`	Evaluate pose without moving atoms	Fastest
`Local only`	Local optimization from starting pose	Fast

The Minimize only mode is particularly valuable for pose refinement workflows, achieving 10-20x speedup over full docking by skipping global search.

Input requirements

Protein (Receptor)

Upload a PDB file or enter an RCSB PDB ID. The protein should be prepared with hydrogens added and missing residues fixed. Use PDB Fixer for automated preparation.

Ligand

Enter a SMILES string for small molecules, or upload an SDF, MOL, or PDBQT file. SMINA automatically generates 3D coordinates and calculates partial charges via OpenBabel.

For complex ligands (macrocycles, peptides, >150 atoms), pre-prepared PDBQT files are recommended to ensure proper handling.

Docking parameters

Scoring function

Default (Vina): Standard Vina scoring, well-validated for general docking
Vinardo: Optimized for pose prediction accuracy, recommended for structure-based drug design
Custom scoring file: Define your own scoring weights for research applications

Docking mode

Full docking: Complete global search with local optimization
Minimize only: Fast pose refinement (10-20x faster than full docking)
Score only: Evaluate a pose without any optimization
Local only: Optimize from the input starting pose

Search parameters

Exhaustiveness: Number of independent docking runs. Higher values (16-32) improve search coverage for difficult targets
Number of poses: Maximum binding poses to return. SMINA supports up to 50 poses, versus Vina's limit of 20
Energy range: Maximum energy difference from the best pose for inclusion. Only poses within this range are returned
RMSD filter: Minimum RMSD between poses to reduce redundancy

Search space

The search box defines where SMINA explores ligand binding:

Auto: Automatically calculates a box encompassing the entire protein surface
Autobox: Derives the search box from a reference ligand (useful when binding site is known from a co-crystal structure)
Manual: Specify exact center coordinates and box dimensions

We recommend using autobox when a reference ligand structure is available, as this focuses the search and improves both speed and accuracy.

Output options

Per-atom interaction energies: Outputs energy contributions for each atom, useful for analyzing which parts of the ligand contribute most to binding

Understanding the results

Binding affinity

Affinity is reported in kcal/mol. More negative values indicate stronger predicted binding:

Range	Interpretation
-4 to -6	Weak binding
-6 to -8	Moderate binding
-8 to -10	Strong binding
< -10	Very strong binding

Vinardo scores may differ slightly from Vina due to the modified scoring function. Both use the same scale and interpretation guidelines.

Pose ranking

Validation metrics

On the CASF-2013 benchmark, Vinardo achieves:

Docking power: 89.7% success rate (pose RMSD < 2Å from crystal structure)
Screening power: Identifies true binders in top 1% for 80% of proteins
Scoring correlation: r = 0.601 between predicted and experimental binding affinities

Comparison to other docking tools

Feature	SMINA	AutoDock Vina	GNINA
Scoring	Vina/Vinardo/Custom	Vina	Vina + CNN
Max poses	50	20	20
Minimization speed	10-20x faster	Standard	Standard
Custom scoring	Yes	No	No
Per-atom energies	Yes	No	No

SMINA is ideal when you need custom scoring functions, fast minimization, or more than 20 poses. For standard docking with known scoring, Vina or GNINA work equally well.

Custom scoring function syntax

Custom scoring functions are defined as weighted sums of interaction terms. Each line specifies a weight and term with parameters:

Text

-0.035579 gauss(o=0,_w=0.5,_c=8)
-0.005156 gauss(o=3,_w=2,_c=8)
0.840245 repulsion(o=0,_c=8)
-0.035069 hydrophobic(g=0.5,_b=1.5,_c=8)
-0.587439 non_dir_h_bond(g=-0.7,_b=0,_c=8)
1.923 num_tors_div

Parameters include:

o: Offset from optimal distance
w: Gaussian width
c: Distance cutoff
g: Good distance threshold
b: Bad distance threshold

Detailed term documentation is available in the SMINA GitHub repository.

Best practices

Use Vinardo for pose prediction: Vinardo consistently outperforms default Vina scoring in docking benchmarks
Define search space when possible: Using autobox with a reference ligand improves both speed and accuracy
Use minimize mode for refinement: When you have a reasonable starting pose, minimize mode is 10-20x faster than full docking
Increase exhaustiveness for difficult targets: Flexible ligands or large binding sites benefit from exhaustiveness values of 16-32