Overview

This field develops models that generate molecular topology and 3D atom coordinates, either from scratch or conditioned on protein pockets, fragments, spectra, or desired properties. Common approaches include SE(3)-equivariant diffusion and flow matching, autoregressive graph generation, latent-space models, and energy-based sampling. Methods are evaluated on benchmarks such as QM9, GEOM-Drugs, and CrossDocked using chemical validity, geometric stability, diversity, property scores, and binding-pose quality. A central challenge is balancing fast, controllable generation with physically stable and diverse molecules that retain realistic distributions and can be synthesized.

Latest papers

CardsList

62 papers5 papers in the last four weeks, down 17% on the four weeks before. 0.0% of all new papers.

  1. Diffusion-based Evolutionary Optimization for 3D Multi-Objective Molecular Generation

    May 16, 2025Ruiqing Sun, Dawei Feng, Sen Yang +53D Molecular GenerationConstrained Generative Modeling

  2. MolMiner: Toward Controllable, 3D-Aware, Fragment-Based Molecular Design

    Nov 10, 2024Raul Ortega-Ochoa, Tejs Vegge, Jes Frellsen3D Molecular GenerationAutoregressive Generation