A review of new science, in plain language

Proceedings of the National Academy of Sciences  ·  5 October 2026  ·  From issue 47

Causality-integrated graph learning for multi-endpoint toxicity prediction

A model using cause-and-effect chemical maps predicts toxic effects better than shape-only models.

Haoyue Tan, Tong Bao et al.

A new computer model predicts how chemicals might harm the body by following known chains of cause and effect, not just by looking at a chemical's shape. It builds a large map linking chemicals to the biological steps they disturb, focused on chemicals that can mess with hormones. In tests, this model stayed more accurate than older top chemical prediction models when facing new chemicals and new situations.

Integrity screen: passed (3 checks) Checked 6 October 2026. Read the source

The plain-language reading of this paper is in the full issue, for subscribers.

Read the full issue

Five vetted papers in plain language, free, every two days

Five papers free every issue, about 15 emails a month. No spam. Leave in one click.

Also in issue 47 15 papers
More from Proceedings of the National Academy of Sciences