Fig. 5

scDALI-Het applied to scRNA-seq of differentiating iPSCs reveals cell-state specificity of eQTL. a, b Q-Q plot of scDALI-Het P-values when permuting the cell-state coordinates of cells from the same donor. A model that does not account for the donor identity yields inflated p-values (a), whereas scDALI-Het with donor identities as fixed effects yields calibrated results (b). c Number of discoveries for varying numbers of principal components (PCs) used to define the cell state kernel. If donor identity is not accounted for, using a larger number of PCs for the cell-state definition leads to an increasing number of discoveries. d Scatter plot of negative log P-values, comparing a model using only the leading PC versus scDALI-Het with 10 PCs. Orange dots are discoveries that are exclusively identified by the general cell-state test (10 PCs). Indicated are the number of significant discoveries in each quadrant (10% FDR). e UMAP visualization of collection- and pseudotime. f Estimated allelic rates for the eGene-QTL pair (CPNE1, chr20:34344225 T/A). g Expression of UTF1, a previously identified marker for neuronal differentiation success