Large Language Models for Zero-shot Inference of Causal Structures in Biology

Izzy Newsham, Luka Kovačević, Richard Moulange, Nan Rosemary Ke, Sach Mukherjee · 2025

Read Large Language Models for Zero-shot Inference of Causal Structures in Biology by Izzy Newsham, Luka Kovačević, Richard Moulange, Nan Rosemary Ke, Sach Mukherjee (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-2503-04347.txt.

Description

Izzy Newsham and co-authors published this computational biology paper in 2025. It tests whether large language models can infer causal links in biology, such as which gene affects which other gene, without any task-specific training. Causal networks in cells are hard to map because effects run through hidden variables and change with context. The team built a framework to check model claims against real experimental data where genes were deliberately perturbed.

The tests covered more than a hundred variables and thousands of causal hypotheses. The authors also tried several ways of prompting the models and of feeding them scientific articles, including large and conflicting sets. With the right setup, even fairly small models captured meaningful parts of the true causal structure. The authors suggest such models could serve as organizers of existing knowledge for later analysis.

Catalog record

CREATORIzzy Newsham, Luka Kovačević, Richard Moulange, Nan Rosemary Ke, Sach Mukherjee
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-04347

Browse the full library · About the Tower of Babel