mlxmolkit 0.1.0

GPU-accelerated molecular 3D coordinate generation on Apple Silicon using MLX

mlxmolkit

GPU-accelerated molecular 3D coordinate generation on Apple Silicon using MLX.

mlxmolkit is a drop-in replacement for RDKit's ETKDG conformer generation that runs on Apple Silicon GPUs via the MLX framework. It matches the interface of NVIDIA's nvMolKit EmbedMolecules, bringing hardware-accelerated molecular embedding to Mac.

Installation

pip install mlxmolkit

Quick Start

from rdkit import Chem
from rdkit.Chem import AllChem
from mlxmolkit import EmbedMolecules

# Prepare molecules
smiles = ["CCO", "c1ccccc1", "CC(=O)Oc1ccccc1C(=O)O"]
molecules = [Chem.AddHs(Chem.MolFromSmiles(s)) for s in smiles]

# Configure ETKDG parameters
params = AllChem.ETKDGv3()
params.useRandomCoords = True

# Generate conformers on GPU
EmbedMolecules(molecules, params, confsPerMolecule=10)

Supports all standard ETKDG variants: ETKDG, ETKDGv2, ETKDGv3, srETKDGv3, KDG, ETDG, and pure DG.

Requirements

• macOS with Apple Silicon (M1/M2/M3/M4)
• Python >= 3.12

Dependencies

• mlx - Apple's ML framework for Apple Silicon
• rdkit - Cheminformatics toolkit
• numpy

License

MIT

Links

• Repository
• Issues