Fig. 2

Illustration of the scGWAS method. A–J Illustration of the normalization process. A The distribution of gene-based p-values calculated from GWAS summary statistics, using bipolar disorder as an example. B Estimation of lambda values in Box-Cox transformation. C QQ plot of the original gene-based p-values from GWAS. D QQ plot of the Box-Cox transformed gene-based p-values. E QQ plot of the Box-Cox transformed and calibrated gene-based p-values. F–J Distribution and QQ plot of the gene expression for the astrocyte cell type in the DER20 panel, in the same order as A–E. K, L Illustration using seven cell types from the DER22 panel to show the impact of the penalty factor, where the top panel shows modules identified including the penalty factor (K) and the bottom panel shows modules identified excluding the penalty factor (L). A full comparison using all cell types can be found in Additional file 1: Fig. S2. M Comparison of different normalization methods for module scores. For each cell type, we show three types of module score distribution: the raw module score, permutation-based z-score, and the z-score based on size-matched random modules from the virtual search process. More comparison examples can be found in Additional file 1: Fig. S3