Fig. 2

Evaluation of correlation metrics. a Comparison of the co-expression profiles among the different single-cell datasets in this study. The Spearman correlation of the Oelen v2 and v3 datasets, the van Blokland v2 and v3 datasets and the van der Wijst dataset were compared with each other, always taking the CD4 + T cells and genes expressed in at least 50% of the cells in the corresponding datasets. The number of genes tested is shown in parentheses below the exact Spearman correlation value. b Comparison of the co-expression profiles between the single-cell datasets and with the bulk RNA-seq datasets from BLUEPRINT, ImmuNexUT (both measuring FACS-sorted naive CD4 + T cells), and BIOS (whole blood). Again, we only assessed genes expressed in at least 50% of the cells for the single-cell dataset (number of tested genes shown in parentheses below the Spearman correlation value). c Relationship between the co-expression similarity between the ImmuNexUT naive CD4 + T cells and Oelen v3 dataset CD4 + T cells and increasing gene expression cutoffs (the ratio of cells with non-zero expression for a given gene). The number of genes tested is indicated by color scale and the numbers in the bar plot. d Comparison of the co-expression profiles between the bulk RNA-seq datasets, taking the same gene subset as in a and b. The number of tested genes is shown in parentheses below the exact Spearman correlation value. e Enrichment of correlated genes in scRNA-seq (Oelen v3 dataset) among associated genes identified by CRISPR knockout. For the enrichment, genes differentially expressed after knockout of FOXP1, FUS, HNRNPK, IRF1, and PCBP1 were identified and tested for enrichment. P-values in the plot show the significance level of the Wilcoxon rank-sum test. f Enrichment of correlated genes in bulk RNA-seq (ImmuNexUT) among associated genes identified by CRISPR knockout. See e and “Methods” for further details