anyone can dock — CLI

Molecular docking from the command line, powered by AutoDock Vina 1.2.7, pKaNET Cloud+, and RDKit.

Installation

System dependencies (Ubuntu / Debian)

sudo apt-get install -y openbabel libopenbabel-dev \
  libcairo2-dev libpango1.0-dev libpangocairo-1.0-0

pip (recommended)

# 1. clone the repository
git clone https://github.com/nyelidl/anyone-docking.git
cd anyone-docking

# 2. install in editable mode  (add [all] for ADMET-AI + PNG export)
pip install -e .
pip install -e ".[all]"

conda

conda create -n acd python=3.11 -y
conda activate acd

# system libs via conda-forge
conda install -c conda-forge openbabel cairosvg rdkit prody -y

# package itself
pip install -e .

After installation the acd command is available in your shell.

Commands

acd --help
acd dock     --help
acd batch    --help
acd codock   --help
acd ensemble --help
acd receptor --help
acd ligand   --help
acd diagram  --help

Quick examples

1. Full single-ligand docking (PDB ID + SMILES)

acd dock \
  --receptor 1M17 \
  --smiles "COCCOC1=C(C=C2C(=C1)C(=NC=N2)NC3=CC=CC(=C3)C#C)OCCOC" \
  --name erlotinib \
  --ph 7.4 \
  --exhaustiveness 16 \
  --output ./erlotinib_dock

Output in ./erlotinib_dock/:

receptor/rec.pdb
receptor/rec.pdbqt
receptor/receptor_summary.json
ligand/erlotinib.pdbqt
erlotinib_out.pdbqt      ← all poses
erlotinib_out.sdf
erlotinib_pv_ready.sdf   ← bond-order corrected
erlotinib_scores.csv

2. Add 2D interaction diagram and save per-pose files

acd dock \
  --receptor 1M17 \
  --smiles "COCCOC1=C(C=C2C(=C1)C(=NC=N2)NC3=CC=CC(=C3)C#C)OCCOC" \
  --name erlotinib \
  --diagram \
  --save-poses \
  --output ./erlotinib_full

3. Use a local PDB/CIF file

acd dock \
  --receptor my_homology_model.pdb \
  --smiles "CCc1nn(C)c2cc(-c3cncnc3Nc3cccc(S(N)(=O)=O)c3)ccc12" \
  --name compound_1 \
  --center manual --cx 14.2 --cy -3.8 --cz 22.1 \
  --bx 22 --by 22 --bz 22

4. Batch docking from a .smi file

# compounds.smi: one "SMILES name" per line
cat > compounds.smi << 'EOF'
O=c1cc(-c2ccc(O)cc2)oc2cc(O)cc(O)c12 Apigenin
O=c1cc(-c2ccccc2)oc2cc(O)c(O)c(O)c12 Baicalein
COCCOC1=C(C=C2C(=C1)C(=NC=N2)NC3=CC=CC(=C3)C#C)OCCOC Erlotinib
EOF

acd batch \
  --receptor 1M17 \
  --ligands compounds.smi \
  --ph 7.4 \
  --exhaustiveness 8 \
  --redock-smiles "COCCOC1=C(C=C2C(=C1)C(=NC=N2)NC3=CC=CC(=C3)C#C)OCCOC Erlotinib" \
  --output ./batch_egfr

Output ./batch_egfr/batch_scores.csv:

Name,SMILES,Prepared SMILES,Top Score,Charge,Status,...
Erlotinib,...,-9.12,0,OK,...
Apigenin,...,-8.44,0,OK,...
Baicalein,...,-8.01,0,OK,...

4a. Simultaneous docking of a ligand set

Use codock when two or more ligands should bind together in one Vina search. This is different from batch, which docks and ranks each ligand independently. The resulting affinity is for the joint ligand complex; it is not a per-ligand score.

acd codock --receptor 5X72 --ligand ligand_p59.pdbqt ligand_p69.pdbqt \
  --center manual --cx -15 --cy 15 --cz 129 --bx 30 --by 24 --bz 24 \
  --exhaustiveness 32 --output ./joint_5x72

--ligand accepts two or more quoted SMILES ("CCO ethanol"), compound names ("aspirin", resolved through PubChem), ligand structure paths (.sdf, .mol2, .pdb, .pdbqt), or a .smi file with one SMILES [name] per line. Inputs can be mixed. The older --ligands, --smiles-list, and --ligand-files options remain available for existing scripts. The output includes joint_out.pdbqt (joint poses) and joint_results.json (ligand labels, files, and joint scores). Check both ligands in each pose before interpreting the result.

5. Prepare receptor once, re-use for many batch runs

# Step A — prepare receptor (saved as JSON)
acd receptor --pdb 1M17 --output ./rec_1m17

# Step B — batch with saved receptor (no re-download or re-prep)
acd batch \
  --receptor-json ./rec_1m17/receptor_summary.json \
  --ligands library.smi \
  --output ./screen_run_1

6. Dock against a receptor ensemble

acd ensemble \
  --receptors "ensemble/*.pdb" \
  --ligands compounds.smi \
  --center manual --cx 14.2 --cy -3.8 --cz 22.1 \
  --bx 22 --by 22 --bz 22 \
  --seed 42 \
  --output ./ensemble_screen

Each receptor is prepared and validated independently, including ferric-heme and Compound-I validation when present. A failed receptor is recorded and skipped without bypassing validation. Use one manual box only for structurally aligned receptors; otherwise use automatic or selection-based centering.

Output in ./ensemble_screen/:

001_model_a/                 complete per-receptor batch output
002_model_b/
ensemble_receptors.csv       preparation and run status for each receptor
ensemble_scores.csv          all receptor-ligand score rows
ensemble_summary.csv         best, mean, median, and population SD per ligand

Ensemble score summaries are ranking aids and are not population-weighted thermodynamic binding energies.

7. Stand-alone 2D diagram for an existing docking result

acd diagram \
  --receptor ./rec_1m17/receptor/rec.pdb \
  --pose-sdf  ./erlotinib_dock/erlotinib_pv_ready.sdf \
  --smiles "COCCOC1=C(C=C2C(=C1)C(=NC=N2)NC3=CC=CC(=C3)C#C)OCCOC" \
  --name erlotinib \
  --pose 1 \
  --score -9.4 \
  --output ./diagrams

Programmatic use (Python API)

from anyonecandock import (
    prepare_receptor,
    prepare_ligand,
    run_vina,
    get_vina_binary,
    fix_sdf_bond_orders,
    draw_interaction_diagram,
)
from pathlib import Path

wdir = Path("./my_run")
wdir.mkdir(exist_ok=True)

# 1. Receptor
rec = prepare_receptor(
    raw_pdb="1M17.pdb",
    wdir=wdir,
    center_mode="auto",
    box_size=(18, 18, 18),
)

# 2. Ligand
lig = prepare_ligand(
    smiles="COCCOC1=C(...)OCCOC",
    name="erlotinib",
    ph=7.4,
    wdir=wdir,
    mode="pkanet",
    conformer_seed=None,  # default: random; set a positive integer to reproduce 3D coordinates
)

# 3. Vina
vina, _ = get_vina_binary()
dock = run_vina(
    receptor_pdbqt=rec["rec_pdbqt"],
    ligand_pdbqt=lig["pdbqt"],
    config_txt=rec["config_txt"],
    vina_path=vina,
    exhaustiveness=16,
    n_modes=10,
    energy_range=3,
    wdir=wdir,
    out_name="erlotinib",
)
print(f"Best score: {dock['top_score']:.2f} kcal/mol")

# 4. 2D diagram
svg = draw_interaction_diagram(
    receptor_pdb=rec["rec_fh"],
    pose_sdf=dock["out_sdf"],
    smiles=lig["prot_smiles"],
    title="Erlotinib · 1M17",
)
Path("diagram.svg").write_bytes(svg)

Environment variables

Optional extras

Citation

If you use Anyone Can Dock in your research, please cite:

Hengphasatporn et al. J. Chem. Inf. Model. 2026 — DFDD
Eberhardt et al. J. Chem. Inf. Model. 2021, 61, 3891–3898 — AutoDock Vina 1.2.7

Ligand structure files

--ligand-file accepts .sdf, .mol2, .pdb, and .pdbqt (case insensitive).

# Prepared PDBQT: original file passed directly to Vina, unchanged
acd dock \
  --receptor-json receptor_summary.json \
  --ligand-file ligand.pdbqt \
  --name ligand \
  --output results

# PDB: existing file preparation and Meeko writer, without adding hydrogen
acd dock \
  --receptor-json receptor_summary.json \
  --ligand-file ligand.pdb \
  --name ligand \
  --no-add-h \
  --output results

PDBQT direct mode performs read-only format checks and verifies the original SHA256 against the file supplied to Vina immediately before launch. It performs no ligand conversion, protonation, hydrogen changes, or atom/type/charge edits. --name controls output filenames only. Missing, empty, malformed, or multi-pose PDBQT inputs fail before docking. Supply one prepared ligand with a valid torsion tree (ROOT, ENDROOT, optional branches, and TORSDOF). Validation checks file structure; Vina remains responsible for engine-specific chemical/type support.

PDB files use prepare_ligand_from_file and the existing _meeko_to_pdbqt implementation (RDKit → Meeko). They do not enter the SMILES prepare_ligand protonation/pKaNET path. Open Babel is not used for ligand PDB → PDBQT conversion, including on failure. Normal PDB mode adds missing hydrogens; --no-add-h disables that addition and requires an input suitable for Meeko, normally with explicit hydrogens already present. For PDB input, this option now also disables Meeko hydrogen merging (merge_these_atom_types=()), preserving the original atom set. Meeko still assigns AutoDock types, Gasteiger partial charges, and torsions. Normal PDB mode retains Meeko’s default nonpolar hydrogen merging.

results/ligand/atom_mapping.tsv records original atom identity, names, residue name/number, chain, insertion code, coordinates, input formal charge, output partial charge/type, and Meeko's explicit atom mapping. Original coordinates are kept, including planar PDB input. New hydrogen atoms are identified separately. In normal mode, nonpolar hydrogens merged by Meeko are reported with their parent; Meeko's torsion-tree reordering is reported without rearranging the PDBQT tree. Original four-column atom-name fields are restored from the input PDB using Meeko’s explicit RDKit-index-to-output-serial mapping, after identity validation. The TSV includes zero-based input_rdkit_index and names before/after restoration. Only the atom-name field is changed by this restoration; charges, types, coordinates, serials, and torsion records remain as Meeko wrote them. Unexpected drops, duplicates, coordinate changes, or ambiguous mapping stop docking. Alternate locations, multiple models, disconnected PDB fragments, and Meeko-generated pseudoatoms are rejected when unique mapping cannot be established. PDB bond-order information is limited: some ligands require a better specified input or externally prepared PDBQT rather than automatic repair.

SDF/MOL2 preparation and their existing Open Babel paths remain available. Open Babel may also still be used for receptor preparation and conversion of docked output PDBQT to SDF. These are separate from ligand input preparation. See PDB atom-name restoration and the implementation audit for source references, validation results, and before/after workflows.