Fig. 6: Integrated genomic and transcriptomic analysis to compare patient tumor versus cell lines.

a Subset analysis of correlation profiling when matching canonical myeloma genomic lesions (three IGH translocations, KRAS/NRAS mutations). Each dot reflects the median Spearman correlation of each cell line carrying the specified genomic lesion correlated versus CoMMpass patients with or without the noted genomic lesion. Box plot shows median and interquartile range. p values by Wilcoxon test. b Heatmap showing patient expression levels of canonical genes (y-axis) overexpressed in each of the seven myeloma subtypes per the analysis of Zhan et al. (“template” = subtype labels as defined by Zhan et al.). Each CoMMpass patient transcriptome was classified into one of the myeloma subtypes based on their relative expression of these canonical genes using the Nearest Template Prediction method (see Methods). False discovery rate (FDR) < 0.05 denotes high-confidence match. Translocations annotated in each sample are also noted and demonstrate close matches to the expected subtypes defined primarily by a chromosomal alteration in Zhan et al. c Similar analysis as in b but for 66 MM cell lines. In this analysis, differential gene expression driving a specific subtype match appears less prominent than in patients. No strong matches are noted for cell line transcriptomes to the HY, LB, and CD-2 subtypes. For b and c, Heatmap scale bar reflects log2-normalized gene expression data with 0 as the median in each row.