Fig. 1
From: Effective methods for bulk RNA-seq deconvolution using scnRNA-seq transcriptomes
We benchmarked data normalizations and deconvolution approaches in datasets with concurrent bulk RNA-seq and scnRNA-seq profiles (*). Cells were clustered in an unsupervised fashion (**). Gold standard abundance estimates (***) for each cell type were obtained by either aggregating cells or nuclei in each scnRNA-seq cluster, immunohistochemistry, fluorescence-activated cell sorting, or cell counts. Deconvolution methods used either full scnRNA-seq expression profiles or cluster-specific biomarkers to predict cell-type abundances based on bulk RNA-seq profiles. Deconvolution accuracies in each sample were assessed by comparing predicted abundances from bulk RNA-seq and gold standard estimates