molex 0.7.3

Parse, analyze, and transform molecular structure data (PDB, mmCIF, BinaryCIF, MRC)

molex

Molecular exchange: a Rust library for parsing, transforming, analyzing, and serializing molecular structure data, with Python and C bindings.

Features

• Parse PDB, mmCIF, BinaryCIF, MRC/CCP4 density maps, and DCD trajectories
• Entity model: proteins, nucleic acids, ligands, ions, waters, and other small molecules as typed entities under a single Assembly
• Analyze: DSSP secondary structure, hydrogen bonds, covalent bonds, disulfide bridges, SASA, contacts
• Transform: Kabsch alignment and superposition RMSD, heavy-atom completion, all-atom projection, structured edits and deltas
• Crystallography (xtal): maximum-likelihood refinement from structure factors: density synthesis, bulk-solvent masking, anisotropic scaling, sigma-A estimation, and B-factor refinement, with an optional GPU backend
• Serialize: a compact binary wire format for FFI and IPC
• Bindings: a PyO3 Python module (Biotite-free numpy interchange) and a C ABI static library (libmolex.a) for embedding in native hosts

Quick start (Rust)

use std::path::Path;
use molex::Assembly;

let assembly = Assembly::from_file(Path::new("1ubq.pdb"))?;
for e in assembly.entities() {
    println!("{:?}: {} atoms", e.molecule_type(), e.atom_count());
}

Assembly::from_file reads PDB or mmCIF by extension; from_pdb / from_mmcif / from_bcif take the source as a string, each with a _with(..., Completion) variant. Heavy-atom completion runs at parse time (default Completion::Heavy). Reach atoms through assembly.entities(), and write back with assembly.to_pdb().

Python

pip install molex

The Python API is object-centric: parse into an Assembly, walk its entities and residues, and read atoms as numpy columns.

import molex

asm = molex.Assembly.from_pdb(open("1ubq.pdb").read())
for e in asm.entities():
    print(e.kind, e.chain_id, e.residue_count)

asm.recompute_ss()   # opt-in DSSP secondary structure

# Biotite-free per-atom numpy columns via PyAtomTable:
table = molex.PyAtomTable.from_assembly_bytes(asm.to_assembly_bytes())
coords = table.coords          # (N, 3) float32
mol_types = table.mol_types    # AtomWorks-style vocabulary columns

from_mmcif / from_bcif parse the other formats. PyAtomTable exposes plain numpy columns (coords, atom/residue names, elements, b-factors, occupancies, chain/residue ids, and AtomWorks-style vocabulary) with no Biotite dependency. Type stubs (molex.pyi) ship with the wheel. Crystallographic refinement is available through molex.PyExperimentalData (from_sf_cif, compute_density, refine_b_factors).

C API

With the c-api feature, molex builds a static library (libmolex.a) plus a cbindgen-generated header (include/molex.h) exposing an opaque-handle C ABI: parse PDB/mmCIF/BinaryCIF into an Assembly, walk entities/residues/atoms, apply edits, and (with xtal) drive crystallographic refinement. This is the interface native hosts embed molex through.

Optional features

Feature	Enables
(default)	Pure-Rust core: parsing, entity model, analysis, wire format
serde	Serialize / Deserialize on the core types
specta	TypeScript type-export derives
python	PyO3 bindings + numpy interchange
extension-module	Build the Python extension as a loadable wheel (with maturin)
c-api	C ABI, libmolex.a, and the generated include/molex.h
xtal	Crystallographic refinement pipeline (density, scaling, sigma-A, FFT)
minimization	B-factor refinement (argmin), on top of xtal
gpu	GPU density/refinement backend (cubecl + wgpu), on top of xtal
testutil	Crystallographic test fixtures for external bench/integration crates

Documentation

• Guide: architecture, modules, and examples
• API reference: generated rustdoc

License

MIT