Fig. 1: Over-accessible distal chromatin classifies myeloma genetic subgroups.

a Study patient population and design. Transcriptome and chromatin accessibility were assessed by RNA-seq and ATAC-seq respectively in myeloma PC from 30 MM patients and PC from three healthy donors. For two of the controls we obtained samples of both CD19⁺and CD19⁻ PC. b Changes in average ATAC-seq signal over pan-myeloma (left) and subgroup (right) peaks expressed as normalized log₂ myeloma/normal PC read count. Significant changes in the pan-myleoma plot (left) are shown in a darker colour. c Logistic regression of number of differentially accessible regions (DAR) within 500 kb of a gene (x-axis), whether a gene has a significantly more open promoter in myeloma (red – yes, black – no) on the probability of the gene being differentially upregulated in myeloma. Points represent fraction of genes upregulated with a given number of DAR within 500 kb and TSS status, and lines represent model fit, grey ribbon 95% confidence interval. Genes with 10 or more DAR were pooled together. d Correlation of signal for differentially accessible ATAC-seq regions and differentially expressed genes within the same topology associating domain (TAD). e Variance explained by the first 17 latent factors (LF) for each of RNA-seq and ATAC-seq signals as calculated by multi-omics factor analysis (MOFA). f, g LF scores for each sample for the first five latent factors in the MOFA model. Subtype for each sample is denoted by colour, as per (a). LF1 and LF2 distinguish normal from MM samples, LF5 distinguishes MAF and CCND1 samples from the rest. h LF scores for a subset of regions and genes used as predictors by MOFA analysis and previously identified as MM-subgroup classifiers¹². The main driving oncogenes MAF, CCND1, CCND2 and NSD2 are highlighted here.