FM-GPT: Bayesian fine mapping for phenome-wide transcriptome-wide association studies.

FM-GPT accurately prioritizes causal genes across multiple phenotypes, reducing candidate sets by up to 80% in phenome-wide TWAS analyses.

• Why it matters: Understanding gene-phenotype relationships is crucial for unraveling complex biological mechanisms and improving disease prediction, but linkage disequilibrium and correlated expressions cause false positives.
• What they did: The authors developed FM-GPT, a Bayesian fine-mapping method that performs gene-guided dimension reduction and handles mixed outcome types across large-scale phenomic data, validated through simulations and UK Biobank applications.
• The result: FM-GPT identified pleiotropic genes affecting brain structure and multiple medical conditions, revealing key biological axes and enabling more precise insights into gene functions and disease mechanisms.