Fig. 8

Systematic error is caused by difference in cell size/cellular RNA content. Experiment schematic: a cell types that were chosen for the experiment, b proportions of the cell type for the experiment. c RNA content for each sequenced sample. d Explained variance by SVD for filtered and unfiltered datasets. Closest 100 genes to each simplex corner were selected to identify cell types. e TSNE plot for the filtered dataset: two clusters can be visually identified. Colored genes are those from the corners. f Gene expression profiles of selected genes in mixed samples and GSEA comparing these genes to differential expression between pure samples of HEK and Jurkat cells. g Proportions predicted by the pipeline. h Bar plot showing relative RNA content to ERCC-spike-in. i Corrected proportions of the pipeline. j Usual pipeline for complete deconvolution: we detect the markers using differential expression and use these markers as input of DSA algorithm. k Proportions predicted by DSA using differential expression markers. l Corrected proportions of k