What it is
Genetic effects on gene expression often vary across cell types, which has motivated mapping expression quantitative trait loci (eQTLs, DNA variants linked to how much a gene is expressed) in single-cell RNA sequencing data. Existing approaches generally aggregate each person's expression across cells (the pseudobulk approach), use Gaussian models poorly suited to read counts, or do not scale. The authors present SAIGE-QTL, a Poisson mixed model that directly models single-cell read counts, enabling calibrated and computationally scalable eQTL mapping while accounting for donor-level and cell-level covariates and unequal numbers of cells per donor. In simulations and three single-cell datasets, it improved statistical efficiency and the discovery of genes with significant eQTLs (eGenes) relative to pseudobulk methods, and produced results consistent with existing single-cell models at substantially lower computational cost. In 1.2 million immune cells from the OneK1K dataset (982 individuals, 14 cell types), it identified 48.8% more eGenes than pseudobulk analysis (17,218 against 11,569, counted per cell type). It also enabled genome-wide trans-eQTL mapping, for variants far from the gene they affect, and set-based tests of rare variants.
Why it matters
In the same immune-cell data, SAIGE-QTL found 48.8% more eGenes than the standard pseudobulk pipeline it was compared with. Part of that gap comes from the gene-level test: given the same ACAT-V combination, the pseudobulk tool TensorQTL found 12,777 eGenes instead of 11,569. SAIGE-QTL also runs set-based tests of rare variants, which the authors say are better powered than single-variant tests for low-frequency variants: 5,541 rare-variant eGenes across the 14 cell types, 2,317 unique genes, of which 483 (21%) remained significant after conditioning on the same genes' common eQTLs, and 413 trans-eQTLs for 390 genes in three cell types at genome-wide significance. Against glmer, which fits the same Poisson mixed model, it cut computation time 244-fold in a small benchmark in one B-cell type, with highly concordant results.
Underlined numbers link to their source. Every metric and quoted figure is listed under Sources and data below.
Filed underGenetic Associations and Epidemiology, Single-cell and spatial transcriptomics, Genetic Mapping and Diversity in Plants and Animals
