MolFORM: Multi-modal Flow Matching for Structure-Based Drug Design
Jie Huang, Daiheng Zhang · 2025
Read MolFORM: Multi-modal Flow Matching for Structure-Based Drug Design by Jie Huang, Daiheng Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-05503.txt.
Description
Jie Huang and Daiheng Zhang present a 2025 model that designs drug-like molecules to fit a protein target's shape. Such structure-based design reads the 3D form of the target and builds molecules that should bind it well. Most top models use diffusion, a slow step-by-step method, while other frame types have had less study.
Their model uses flow matching on two modes at once: the atom types, which are discrete, and the atom places, which are continuous. A fine-tune stage then steers the model with a reward based on a docking score, using a preference method. The steer acts on both modes in step, which keeps them in tune. Scores rose across many test measures after the steer. Code for the model is shared in public.
Catalog record
| CREATOR | Jie Huang, Daiheng Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-05503