We introduce a new graph diffusion model for small molecule generation, \emph{DMol}, which outperforms the state-of-the-art DiGress model in terms of validity by roughly across all benchmarking datasets while reducing the number of diffusion steps by at least -fold, and the running time to roughly one half. The performance improvements are a result of a careful change in the objective function and a ``graph noise" scheduling approach which, at each diffusion step, allows one to only change a subset of nodes of varying size in the molecule graph. Another relevant property of the method is that it can be easily combined with junction-tree-like graph representations that arise by compressing a collection of relevant ring structures into supernodes. Unlike classical junction-tree techniques that involve VAEs and require complicated reconstruction steps, compressed DMol directly performs graph diffusion on a graph that compresses only a carefully selected set of frequent carbon rings into supernodes, which results in straightforward sample generation. This compressed DMol method offers additional validity improvements over generic DMol of roughly , increases the novelty of the method, and further improves the running time due to reductions in the graph size.
View on arXiv@article{niu2025_2504.06312,
title={ DMol: A Schedule-Driven Diffusion Model for Highly Efficient and Versatile Molecule Generation },
author={ Peizhi Niu and Yu-Hsiang Wang and Vishal Rana and Chetan Rupakheti and Abhishek Pandey and Olgica Milenkovic },
journal={arXiv preprint arXiv:2504.06312},
year={ 2025 }
}