Fig. 6
From: Complete deconvolution of cellular mixtures based on linearity of transcriptional signatures
Complete deconvolution pipeline is able to dissect realistic datasets. a Pipeline steps required to perform complete deconvolution. b Schematic of GSE27563 dataset. c First five components of SVD of the filtered dataset explain more that 97% of the variance. Closest 100 genes to each simplex corner for panels d and e. d TSNE dimensionality reduction can be used to highlight the structure of high-dimensional filtered dataset. e We used datasets GSE27787 for mouse hematopoietic cells, and GSE49664 for erythrocyte-related population to identify cell types in the mixture. Heatmaps show averaged z-score of identified gene sets in population presented in the datasets. Values are from gene set minimum (blue) to gene set maximum (red). f Box and dot plot for deconvolution results, showing changes between normal and tumor-bearing mice. For each boxplot bottom whisker, bottom of the box, middle line, top of the box, and top whisker are 5%, 25%, 50% (median), 75%, and 95% quantiles respectively. Group comparisons were determined using a two-sided Mann–Whitney U test (n.s. p value >0.05, *p value <0.05, ***p value <0.001, ****p value <0.0001)
