Soft causal learning for generalized molecule property prediction: An environment modeling perspective

Limin Li, Kuo Yang, Wenjie Du, Pengkun Wang, Zhengyang Zhou, Yang Wang · 2025

Read Soft causal learning for generalized molecule property prediction: An environment modeling perspective by Limin Li, Kuo Yang, Wenjie Du, Pengkun Wang, Zhengyang Zhou, Yang Wang (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-2505-06283.txt.

Description

Limin Li and five co-authors present a 2025 arXiv paper on predicting molecule properties for compounds unlike any in the training set. Drug models usually learn from graphs of atoms, but new chemical environments break the patterns they memorized. Earlier fixes hunted for a stable core pattern inside each molecule, yet real chemistry outgrows that idea as environments expand.

Their framework, soft causal learning, models the environment itself instead. A generator grows molecule graphs in line with chemistry rules to imitate new surroundings. An objective then teases environment apart from the molecule's stable part, and an attention step lets the two interact softly rather than by rigid cuts. Across seven data sets staged as different kinds of shift, the method generalized better than rivals. The authors frame it as a causal view of why molecules behave as they do.

Catalog record

CREATORLimin Li, Kuo Yang, Wenjie Du, Pengkun Wang, Zhengyang Zhou, Yang Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-06283

Browse the full library · About the Tower of Babel