Predicting function of evolutionarily implausible DNA sequences

Shiyu Jiang, Xuyin Liu, Zitong Jerry Wang · 2025

Read Predicting function of evolutionarily implausible DNA sequences by Shiyu Jiang, Xuyin Liu, Zitong Jerry 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-2506-10271.txt.

Description

Genomic language models learn from natural DNA, raising a question: do they understand what DNA does, or do they just recognize familiar patterns? Shiyu Jiang, Xuyin Liu, and Zitong Jerry Wang built a benchmark called Nullsettes to test that gap. It uses synthetic expression cassettes with little evolutionary precedent. Models must predict which mutations shut a cassette's function off, even when the sequence looks implausible by evolutionary standards.

Twelve leading models took the test, and most failed to spot the strong loss-of-function mutations reliably. Every model's accuracy fell sharply as the original sequence grew less likely under its own scoring. The authors conclude the models lean on their evolutionary prior instead of a working grasp of gene expression. They call for benchmarks and training that reward functional understanding.

Catalog record

CREATORShiyu Jiang, Xuyin Liu, Zitong Jerry Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-10271

Browse the full library · About the Tower of Babel