Large language models identify causal genes in complex trait GWAS

Identifying underlying causal genes at significant loci from genome-wide association studies (GWAS) remains a challenging task. Literature evidence for disease-gene co-occurrence, whether through automated approaches or human expert annotation, is one way of nominating causal genes at GWAS loci. However, current automated approaches are limited in accuracy and generalizability, and expert annotation is not scalable to hundreds of thousands of significant findings. Here, we demonstrate that large language models (LLMs) can accurately identify genes likely to be causal at loci from GWAS. By evaluating the performance of GPT-3.5 and GPT-4 on datasets of GWAS loci with high-confidence causal gene annotations, we show that these models outperform state-of-the-art methods in identifying putative causal genes. These findings highlight the potential of LLMs to augment existing approaches to causal gene discovery. ### Competing Interest Statement S.S., W.W., S.K., X.W., A.R., A.A., A.K., are employed by and hold stock or stock options in 23andMe, Inc. ### Funding Statement This study did not receive any funding ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: All source data were openly available. Download links: * OpenTargets - https://github.com/opentargets/genetics-gold-standards/ * Pharmaprojects - https://github.com/ericminikel/genetic_support * Weeks et al. - https://www.finucanelab.org/data * GWAS Catalog - https://www.ebi.ac.uk/gwas/docs/file-downloads I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes We will share all processed datasets used in our analysis, as well as the prediction results from all methods on all datasets, intermediate outputs like gene and phenotype embeddings using Zenodo (doi: 10.5281/zenodo.11391053).