Fig. 2

Results of CDSeqR analysis of the in vitro cell mixtures. The mixtures comprise 4 four cell types: cancer-associated fibroblast, lymphocytes, normal breast cells, tumor cells (see Additional file 1: Methods for details). We ran CDSeqR on 32 bulk RNA-seq samples with known cell proportions in each sample [5] using 7 randomly sampled gene subsets of size 500 (7 data blocks, each of 500 genes). In the first row, each point represents a sample (n = 32); in the second row, each point represents a gene (n =  ~ 19,000). True and estimated GEPs are normalized as RPKM. The line in each panel is the x = y line