Extended Data Fig. 5: Concordance of pseudobulk and single-cell genotype effect sizes.
From: Precisely defining disease variant effects in CRISPR-edited single cells
(a) Genotype effect sizes from pseudobulk and single-cell eQTL analysis were compared using Deming regression to estimate a scaling beta. (b) Z-score statistics were computed using the likelihood ratio test (LRT) p-value from the single-cell model or the nominal p-value from the pseudobulk model. The LRT p-value compares the null model (without genotype) to the full model (including genotype). Concordance between Z-scores from the single-cell and pseudobulk models was measured using Pearson’s correlation (r). (c) The violin plots represent log(CP10K + 1)-normalized single-cell expressions grouped by genotype dosage for a given gene (N = 445,790 cells), while the box plots show log2(CPM + 1)-normalized pseudobulk expressions per donor grouped by dosage for a given gene (N = 259 individuals). Boxes show the interquartile range (IQR), whiskers show the 1.5*IQR.
