Fig. 5
From: Effective methods for bulk RNA-seq deconvolution using scnRNA-seq transcriptomes
Mixture deconvolution with transformed RNA-seq data. A DWLS deconvolved the composition of our mixtures with near-perfect accuracy when given the bulk RNA-seq expression profiles of each cell type (r = 0.98, RMSE = 0.04) and B with high accuracy when using cell-type expression estimates from scRNA-cluster profiles (r = 0.93, RMSE = 0.08). C SQUID deconvolution accuracy, relative to cell counts, when using cell-type expression estimates from scRNA-cluster profiles (r = 0.95, RMSE = 0.06) was significantly better than DWLS (p < 2E − 4, Fisher’s transformation). D Deconvolution accuracies of concurrent RNA-seq and scnRNA-seq profiled tissues using SQUID, DWLS, and OLS, as assessed by Pearson correlation and root mean square error (RMSE)