Transcriptome-wide association studies at cell state level using single-cell eQTL data

Area of Science:

• Genetics
• Computational Biology
• Immunology

Background:

• Transcriptome-wide association studies (TWAS) are crucial for identifying genes linked to diseases.
• Existing TWAS methods often overlook crucial cellular heterogeneity by analyzing bulk or pseudobulk tissues.
• This limitation hinders the precise understanding of gene-disease associations within complex biological systems.

Purpose of the Study:

• To introduce TWiST, a novel statistical method for TWAS that operates at single-cell resolution.
• To leverage single-cell expression quantitative trait loci (eQTL) data for more granular genetic association analyses.
• To enable the detection of dynamic and nonlinear gene effects across cell states.

Main Methods:

• TWiST utilizes pseudotime to define and analyze distinct cell states.
• The method models gene expression effects on traits as continuous pseudotemporal curves.
• It allows for flexible hypothesis testing, including global, dynamic, and nonlinear associations.

Main Results:

• Simulations and real-world data analyses confirm TWiST's superior statistical power over traditional pseudobulk methods.
• TWiST effectively captures cell state heterogeneity, leading to more accurate gene prioritization.
• Application to the OneK1K study revealed numerous genes with dynamic effects on autoimmune diseases during immune cell differentiation.

Conclusions:

• TWiST significantly enhances the power of TWAS by incorporating cell state information from single-cell eQTL data.
• The method provides a powerful framework for dissecting complex genetic architectures of diseases.
• TWiST holds substantial promise for advancing our understanding of disease genetics through single-cell analyses.