A Pseudotime-Dependent TWAS Framework Identifies Disease Genes along Cell Developmental Paths.

Transcriptome-wide association studies (TWAS) link genes to disease risk by integrating gene expression with genome-wide association study (GWAS) data, where the use of bulk-tissue expression data typically provides gene-disease association interpretations at tissue levels. Recently, the increasing availability of single-cell gene expression profiles provides an opportunity to to dissect these associations at finer cellular granularity, allowing identification of cell-level effects that are not discernible from bulk-level analyses. While existing methods leverage single-cell data and map associations into discrete cell types, they may miss the continuous nature of cellular processes and misidentify causal cell stages in which genes exert their effects. To capture these continuous dynamic changes in gene expression, we developed the pseudotime-dependent Transcriptome-wide association study (pt-TWAS), a novel TWAS framework that captures gene effects along cell developmental paths and reveals their associations at a finer cell-stage resolution. By modeling gene expression as a continuous function of pseudotime, pt-TWAS gains statistical advantages over methods analyzing discrete cell types or stages. Specifically, it boosts statistical power by borrowing expression quantitative trait loci (eQTL) information across cell stages and jointly testing the gene-disease associations. Furthermore, pt-TWAS constructs and visualizes simultaneous confidence bands for the gene effect curve to identify the causal cell stage for the disease. As a demonstration of our method, we applied pt-TWAS to a GWAS of B-cell acute lymphoblastic leukemia (ALL) leveraging single-cell data from OneK1K, where we successfully replicated known risk genes from previous analyses and pinpointed their relevant cell stages. An R package implementing pt-TWAS is available at https://github.com/RuiCao34/ptTWAS/ .