dockq-rs 0.1.0

Fast Rust reimplementation of the DockQ protein/nucleic-acid docking scorer, with a Python API

dockq-rs

Fast Rust reimplementation of the DockQ protein / nucleic-acid docking scorer, with a Python API. Same scores as DockQ v2.1.3 (validated to f32 precision against the reference), ~20–25× faster, with parallel batch scoring.

pip install dockq-rs

Usage

import dockq_rs

# single model vs native
r = dockq_rs.score("model.pdb", "native.pdb")
print(r["GlobalDockQ"], r["best_mapping_str"])

# rank many models against one native (parallel in Rust)
outcomes = dockq_rs.score_one_vs_many("native.pdb", ["m1.pdb", "m2.pdb", "m3.pdb"])

# arbitrary (model, native) pairs (parallel)
outcomes = dockq_rs.score_pairs([("m1.pdb", "n1.pdb"), ("m2.pdb", "n2.pdb")])

Migrating from DockQ

The function-level API mirrors the reference. Change:

from DockQ.DockQ import load_PDB, run_on_all_native_interfaces

to:

from dockq_rs import load_PDB, run_on_all_native_interfaces

load_PDB(path, chains=[], small_molecule=False, n_model=0) and run_on_all_native_interfaces(model, native, chain_map=..., no_align=False, capri_peptide=False) keep the same arguments and return shapes — the per-interface dict has the same keys (DockQ, F1, iRMSD, LRMSD, fnat, fnonnat, nat_correct, nat_total, nonnat_count, model_total, clashes, len1, len2, class1, class2, chain1, chain2).

Scope

Protein and nucleic-acid scoring, all mapping modes, capri_peptide, no_align, allowed_mismatches. Small-molecule (HETATM) symmetry-corrected scoring is not implemented in this build and raises a clear error (no silent fallback).

Links

Source, standalone CLI, benchmarks, and correctness details: https://github.com/aarteixeira/dockq-rs. MIT licensed.