PLOS Genetics study introduces FM-GPT, a Bayesian fine-mapping method for large-scale transcriptome-wide association studies

A study published in PLOS Genetics introduces FM-GPT, a Bayesian fine-mapping method designed to address a persistent challenge in transcriptome-wide association studies (TWAS): that linkage disequilibrium and correlated gene expression produce spurious signals that are difficult to distinguish from genuine gene–trait associations.

TWAS integrates genome-wide association study (GWAS) data with expression quantitative trait locus (eQTL) reference panels to identify genes associated with traits or diseases. However, as multiple genes in a region tend to have correlated expression patterns driven by shared regulatory variants, the approach frequently implicates more genes than are truly causal. Fine mapping — narrowing down which signals within a locus are most likely to be driving an association — is therefore essential for translating TWAS results into testable biological hypotheses.

FM-GPT extends fine-mapping to phenome-wide settings, enabling simultaneous analysis across the large trait collections now available in electronic health record (EHR) biobanks. The authors, based at institutions including the University of Maryland and the University of Central Florida, demonstrate the method on publicly available phenomic resources, showing improved prioritisation of putatively causal genes relative to existing approaches.

The method is likely to be of interest to statistical geneticists working with large EHR-linked biobanks and to translational researchers using TWAS to nominate drug targets.