An embedding-based framework enables statistical testing of gene-set function hypotheses inferred by large language models
Yanhao Tan, Li-Ju Wang, Tianyuzhou Liang, Ying-Ju Lai, Chien-Hung Shih, Yibing Guo, Tyler M Yasaka, George C Tseng, Yu-Chiao Chiu · 2026
Read An embedding-based framework enables statistical testing of gene-set function hypotheses inferred by large language models by Yanhao Tan, Li-Ju Wang, Tianyuzhou Liang, Ying-Ju Lai, Chien-Hung Shih, Yibing Guo, Tyler M Yasaka, George C Tseng, Yu-Chiao Chiu (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/PMC13526818.txt.
Description
Tan and colleagues published a 2026 study on checking AI guesses about gene jobs with hard statistics. Language models can now guess a gene set's role from its list alone. Those guesses sound plausible but lacked a rigorous way to be tested. The team turned gene and role descriptions into number vectors for testing.
Top embedding models recovered known gene relationships with high accuracy overall. They caught gene-role links with sensitivity above 95 percent in benchmarks. The framework separated real signals from model-made noise better than rivals. An open software package puts the method in other labs' hands. One drug-response gene set gained meaning through the new approach. Language-model guesses thus gain a proper gatekeeper for research use.
Catalog record
| CREATOR | Yanhao Tan, Li-Ju Wang, Tianyuzhou Liang, Ying-Ju Lai, Chien-Hung Shih, Yibing Guo, Tyler M Yasaka, George C Tseng, Yu-Chiao Chiu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 85 KB |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |