2505.00518

Total: 1

#1 An evaluation of unconditional 3D molecular generation methods [PDF] [Copy] [Kimi] [REL]

Authors: Martin Buttenschoen, Yael Ziv, Garrett M. Morris, Charlotte M. Deane

Unconditional molecular generation is a stepping stone for conditional molecular generation, which is important in \emph{de novo} drug design. Recent unconditional 3D molecular generation methods report saturated benchmarks, suggesting it is time to re-evaluate our benchmarks and compare the latest models. We assess five recent high-performing 3D molecular generation methods (EQGAT-diff, FlowMol, GCDM, GeoLDM, and SemlaFlow), in terms of both standard benchmarks and chemical and physical validity. Overall, the best method, SemlaFlow, has a success rate of 87% in generating valid, unique, and novel molecules without post-processing and 92.4% with post-processing.

Subjects: Chemical Physics , Quantitative Methods

Publish: 2025-05-01 13:40:10 UTC