chematic 0.4.15

Pure-Rust cheminformatics — SMILES, fingerprints, 70+ descriptors, pKa, ADMET

chematic

Pure-Rust cheminformatics library for Python — SMILES parsing, 70+ molecular descriptors, fingerprints, pKa prediction, ADMET profiling, and template-based retrosynthesis.

Installation

pip install chematic

Quick Start

import chematic

mol = chematic.from_smiles("CC(=O)Oc1ccccc1C(=O)O")  # aspirin

print(mol.mw)      # 180.16
print(mol.logp)    # 1.31
print(mol.tpsa)    # 63.6
print(mol.qed)     # 0.55

# pKa prediction
print(mol.pka())   # {"most_acidic": 3.49, "most_basic": None}

# ADMET profile
print(mol.admet())
# {"bbb": False, "bbb_score": ..., "caco2": ..., "herg_risk": ..., "cyp3a4_risk": ...}

# Fingerprints (bytes, 2048-bit ECFP4)
fp = mol.ecfp4()

# Tanimoto similarity
mol2 = chematic.from_smiles("c1ccccc1")
sim = chematic.tanimoto(mol.ecfp4(), mol2.ecfp4())

# Template-based retrosynthesis (60 retro-SMIRKS templates)
mol3 = chematic.from_smiles("CC(=O)Nc1ccccc1")  # acetanilide
results = mol3.retro_disconnect(max_results=5)
for r in results:
    print(r["template"], "→", r["precursors"])
# amide_secondary → ['CC(=O)O', 'Nc1ccccc1']

# Filter by reaction class
amides = mol3.retro_disconnect(reaction_class="AmideBond")

# All descriptors as a dict (for Pandas)
import pandas as pd
smiles = ["CCO", "c1ccccc1", "CC(=O)O"]
df = pd.DataFrame([chematic.from_smiles(s).descriptors() for s in smiles])

Features

• Zero C/C++ dependencies — pure Rust, no RDKit or OpenBabel required
• SMILES / MOL / SDF parsing and writing
• 70+ descriptors: MW, LogP (±0.3 vs RDKit), TPSA (±1.0 Ų), QED, Fsp3, SA Score, HBD (100% vs RDKit, incl. S-H)
• 14 fingerprint algorithms: ECFP2/4/6, FCFP4/6, MACCS, AtomPair, Torsion, …
• pKa prediction (15 SMARTS rules — unique to chematic)
• ADMET profile: BBB, Caco-2, hERG, CYP3A4
• Template-based retrosynthesis: mol.retro_disconnect() — 60 retro-SMIRKS templates, SA Score ranked
• SMARTS substructure search
• SVG depiction for Jupyter notebooks

License

MIT OR Apache-2.0