Fig. 6

Effect of background removal on downstream analysis. A UMAP representation of replicate 2 single-cell data before and after background noise correction, colored by cell type labels obtained from reference based classification. Individual cells that received a new label after correction are highlighted. PT, proximal tubule; CD_IC, intercalated cells of collecting duct; CD_PC, principal cells of collecting duct; CD_Trans, transitional cells of collecting duct; CNT, connecting tubule; DCT, distal convoluted tubule; Endo, endothelial; Fib, fibroblasts; aLOH, ascending loop of Henle; dLOH, descending loop of Henle; MC, mesangial cells; Podo, podocytes. B Expression of the PT cell marker Slc34a1 before and after background noise correction in replicate 2. Cells that were classified as PT cells in the uncorrected data, but got reassigned after correction, are highlighted. C, D Differential expression analysis of 10 PT markers, evaluating the expression fraction in non-PT cells (C) and the log2 fold change between PT and all other cells (D). E Evaluation metrics for the effect of background noise correction on classification and clustering. For each metric the change relative to the uncorrected data is depicted. The values were scaled by the possible range of each metric. Prediction score: cell-wise score “delta” of reference based classification with SingleR [20]. Average silhouette: Mean of silhouette widths per cell type. Purity: Cluster purity calculated on cell type labels as ground truth and Louvain clusters as test labels. k-NN overlap: overlap of the k=50 nearest neighbors per cell compared to genotype-cleaned reference k-NN graph