Efficient count-based models improve power and robustness for large-scale single-cell eQTL mapping.

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) allows variant effect characterization on gene regulation at the cellular level (single-cell eQTLs; sc-eQTLs).
• Existing sc-eQTL mapping methods struggle with sparse scRNA-seq counts and large datasets.

Purpose of the Study:

• To develop a flexible and efficient framework for sc-eQTL mapping using count-based models.
• To improve the identification of eQTLs, particularly those missed in bulk tissue analysis.

Main Methods:

• Proposed jaxQTL, an efficient sc-eQTL mapping framework utilizing count-based models on pseudobulk data.
• Validated jaxQTL using extensive simulations and applied it to the OneK1K scRNA-seq dataset (N=982) across 14 cell types.

Main Results:

• jaxQTL with a negative binomial model outperformed other methods in identifying sc-eQTLs with calibrated type I error.
• Identified 11-16% more eGenes than existing approaches, primarily due to detecting lowly expressed genes.
• Fine-mapped sc-eQTLs were located further from transcription start sites (TSS) and enriched in cell-type-specific enhancers compared to bulk-eQTLs.
• sc-eQTLs explained more SNP-heritability for 16 blood/immune traits than bulk-eQTLs.
• Demonstrated jaxQTL's utility by nominating IL6ST as a candidate gene for rheumatoid arthritis using T cell sc-eQTLs.

Conclusions:

• jaxQTL offers an efficient and powerful approach for identifying missing disease-associated eQTLs using count-based models.
• sc-eQTLs provide valuable insights into distal regulatory elements and genetic contributions to disease, complementing bulk tissue analyses.