Key points are not available for this paper at this time.
Recent advances in transcriptome-wide association study (TWAS) fine-mapping have enabled the joint modeling of multiple genes to improve causal gene prioritization. However, existing methods have been developed primarily for quantitative traits and most of them rely on gene expression data from a single tissue. Here, we present mFABIO, a multi-tissue TWAS fine-mapping method specifically designed for binary traits. mFABIO employs a probit model to directly link genetically regulated expression (GReX) of genes within a locus across multiple tissues to a binary outcome, while accounting for correlations in GReX across genes and tissues. As a result, mFABIO offers substantial power gains for binary traits, while maintaining robust control of false discovery rates (FDR). We evaluated mFABIO through extensive simulations and applied it to an in-depth analysis of six binary disease traits (asthma, breast cancer, gout, hypertension, prostate cancer, and rheumatoid arthritis) in the UK Biobank, using expression data spanning 38 Genotype-Tissue Expression (GTEx) tissues. mFABIO identified an average of 42 likely causal genes and 65 tissue-gene pairs per disease (FDR < 0.05). Notably, 60.9% of the genes and 77.2% of the gene-tissue pairs were supported by existing TWAS or GWAS evidence. This represented at least a 14.9% increase in evidence-supported genes and a 14.8% increase in evidence-supported gene-tissue pairs, compared to existing approaches. Additionally, mFABIO was also able to narrow down the list of potentially causal candidates by at least 51.3% for genes, and 50.8% for gene-tissue pairs, compared to single-tissue approaches. Leveraging its improved power, mFABIO successfully prioritized multiple potentially causal gene-tissue pairs associated with these diseases, with biological support. Notable examples include D2HGDH in lung tissue for asthma, CYBRD1 in breast mammary tissue for breast cancer, and CCR6 in spleen tissue for rheumatoid arthritis. Overall, mFABIO serves as an effective tool for multi-tissue TWAS fine-mapping of binary traits.
Zhang et al. (Wed,) studied this question.