Fig. 3

Comparisons with existing deconvolution methods using data generated from benchmark pipeline [17]. We compared CDSeqR with 14 other deconvolution methods using synthetic mixtures generated from three scRNA-seq datasets [19, 20, 40]. The results for the PBMC dataset [40] are shown here. The X-axis represents different deconvolution methods whereas the Y-axis denotes the correlation between the estimated cell type proportions by each method and the true cell type proportions. We generated 10 sets of 100 mixtures. Each dot in the box plots represents a deconvolution result from one of the 10 sets. CDSeqRR refers to deconvolution with the Reduce-Recover strategy whereas CDSeq denotes deconvolution without using Reduce-Recover option Additional results of the comparisons are available in Additional file 1: Figs. S6–S8