ActivityDiff: a diffusion model with positive and negative activity guidance for de novo drug design

Huimin Zhu, Renyi Zhou, Jing Tang, Min Li, Daisuke Kihara · 2026

Read ActivityDiff: a diffusion model with positive and negative activity guidance for de novo drug design by Huimin Zhu, Renyi Zhou, Jing Tang, Min Li, Daisuke Kihara (2026) 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/PMC13478866.txt.

Description

Zhu and colleagues introduce a generative model that designs drug molecules while avoiding unwanted targets in this 2026 paper. Most molecule-generating systems chase activity at the intended target and ignore off-target binding, which causes side effects and safety failures. Their system, ActivityDiff, is a diffusion model guided by two separately trained classifiers, one scoring desired activity and one scoring activity to steer away from.

The dual guidance let the model handle several design jobs. It generated molecules for one or two targets at once. It designed around a fixed molecular fragment. And it sharpened selectivity for a chosen target. It also reduced off-target effects in the reported experiments compared with positive-only approaches. The authors argue that building avoidance into generation itself is a practical route to safer, more selective candidate drugs.

Catalog record

CREATORHuimin Zhu, Renyi Zhou, Jing Tang, Min Li, Daisuke Kihara
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13478866

Browse the full library · About the Tower of Babel