chematic-chem

Pure Rust chemical intelligence library — descriptors, drug-likeness rules, diversity picking, reaction analysis.

Features

Molecular Descriptors (40+)

Physicochemical: MW, LogP (Crippen, ±0.3 vs RDKit on 175 mol), TPSA (±1.0 Å² vs RDKit on 175 mol), PSA, MOLAR_REFR, VdW volume
Van der Waals volume: vabc(mol) -> f64 — Bondi radii + spherical-cap overlaps, no 3D coordinates required
Lipophilicity: LogP (multiple models), MolLogP
Hydrogen Bonding: HBA (99.98% RDKit agreement on 5 000 molecules), HBD (100% agreement, including S-H thiols), HBA_LIPINSKI, HBD_LIPINSKI
Flexibility: Rotatable bonds, ring count, scaffold RMSD
Complexity: Topological Polar Surface Area, QED, SA Score
Topological indices:
- schultz_mti(mol) -> u64 — Schultz Molecular Topological Index: Σ_{i<j}(δᵢ+δⱼ)×dᵢⱼ
- gutman_mti(mol) -> u64 — Gutman MTI*: Σ_{i<j}δᵢ×δⱼ×dᵢⱼ
- gravitational_index(mol) -> f64 — Σ_{i<j}mᵢmⱼ/dᵢⱼ² (atomic masses, graph distance)
Molecular Properties: Atom/bond counts, isotopes, fragments
Fingerprint Similarity: Tanimoto, Dice (ECFP, MACCS)

Drug-Likeness Rules

Lipinski Rule of Five: MW < 500, LogP < 5, HBA ≤ 10, HBD ≤ 5
Veber's Rules: TPSA ≤ 140, RotBonds ≤ 10
Egan's Rules: LogP 0.4-5.0, TPSA 20-130
Ghose's Rule: MW 160-480, LogP -0.4-5.6, TPSA 40-130
REOS Filters: Removes likely PAINS, toxic patterns
PAINS Filters: Pan-Assay Interference compounds + alerts
Brenk Filters: Toxicity / instability structural alerts
BOILED-Egg (boiled_egg() → BoiledEggProfile): GI absorption + BBB penetration prediction (Daina & Zoete 2016)

Advanced Analysis

Murcko Scaffold: Generic scaffold from molecule
Maximum Common Substructure (MCS): Molecular similarity
Reaction Analysis: Atom-map detection, reaction center identification
Stereochemistry: Invert/enumerate stereoisomers
Isotope Distribution: Exact mass spectrum simulation

Diversity Selection

MaxMin Picking: Diverse subset selection
Butina Clustering: Distance-based clustering
Tanimoto Similarity: ECFP-based distances

Standardization

Mol Hash: Canonical identifier (accounting for salts, tautomers)
Standardize: Remove/identify salts, neutralize charges
Parse Condensed: IUPAC-like condensed formulas

Quick Start

Calculate descriptors

use chematic_chem::molecular_weight;
use chematic_smiles::parse;

let mol = parse("CC(=O)Oc1ccccc1C(=O)O")?; // aspirin
let mw = molecular_weight(&mol);
println!("MW: {:.2}", mw);

Check drug-likeness

use chematic_chem::{
    lipinski_descriptor_pass,
    veber_descriptor_pass,
    pains_filters
};

if lipinski_descriptor_pass(&mol) {
    println!("✓ Passes Lipinski's Rule of Five");
} else {
    println!("✗ Violates Lipinski");
}

let pains = pains_filters(&mol);
if !pains.is_empty() {
    println!("PAINS alerts: {:?}", pains);
}

Similarity search

use chematic_chem::compare_molecules;

let results = compare_molecules(&[smiles1, smiles2])?;
for pair in &results.pairwise {
    println!("Similarity: {:.3}", pair.similarities.ecfp4_tanimoto);
}

Diversity picking

use chematic_chem::maxmin_diversity;

let picks = maxmin_diversity(&molecules, 10)?;
println!("Selected {} diverse molecules", picks.len());

API Reference

Function	Purpose
`molecular_weight(mol)`	Average isotope mass
`logp(mol)`	XLogP model
`tpsa(mol)`	Topological Polar Surface Area
`vabc(mol)`	Van der Waals volume (Bondi radii, no 3D needed)
`qed(mol)`	Drug-likeness (0-1)
`sa_score(mol)`	Synthetic accessibility (1-10)
`schultz_mti(mol)`	Schultz Molecular Topological Index
`gutman_mti(mol)`	Gutman MTI*
`gravitational_index(mol)`	Gravitational index (Σ mᵢmⱼ/dᵢⱼ²)
`lipinski_descriptor_pass(mol)`	Lipinski's Rule of Five
`compare_molecules(smiles_list)`	Multi-mol similarity
`screen_smiles(smiles_list)`	Batch descriptor + filtering
`maxmin_diversity(mols, k)`	Diverse subset picking
`find_mcs(mols)`	Maximum common substructure
`boiled_egg(mol)`	BOILED-Egg GI/BBB prediction (→ `BoiledEggProfile`)

Crate Dependencies

chematic-core — Atom, Bond, Molecule types
chematic-smiles — SMILES parsing
chematic-perception — Ring, aromaticity, stereo
chematic-fp — Fingerprint calculation
chematic-mol — File I/O

Zero FFI: Pure Rust, WASM-compatible via npm @kent-tokyo/chematic.

Testing

cargo test --lib
# 496 tests: descriptors, rules, MCS, diversity

Version History

v0.1.94 (2026-06-12):

SA Score fragment corpus expanded: 145 → 188 FDA molecules (1034 → 1415 unique fragments)
Integrated with multi-sphere CIP and enhanced fingerprints

v0.1.93 (2026-06-12):

Full multi-sphere CIP priority rules (moved from chematic-chem to chematic-perception)
Correct R/S stereochemistry assignment (>2 distinct substituents)

v0.1.32 (2026-06-07):

992 total tests (v0.1.30: 948)
Integrated with v0.1.32 crate updates (3D constraints, aromaticity)

License

MIT OR Apache-2.0

chematic-chem 0.4.24