Extended Data Fig. 3: SCAVENGE identifies trait-relevant cell populations in a real scATAC-seq dataset.
From: Variant to function mapping at single-cell resolution through network propagation
a, UMAP plots illustrate the canonical marker genes of monocytes (S100A12, MPO), B cells (PAX5, MS4A1) and T cells (CD7, CD3D). The cells are colored by corresponding gene activity score. b, Distribution of network propagation scores before and after scaling and normalization. SCAVENGE analyses for monocyte count on the 10X PBMC dataset are performed. The distributions are in high concordance before and after scaling and normalization. c, Evaluation of SCAVENGE’s robustness to parameter choice. SCAVENGE analyses are performed with (left) different numbers of k (number of nearest neighbors) used for cell-to-cell network construction and (right) different fractions of cells selected as seed cells. The resulting TRSs shown in the boxplots are robust to different parameters and can rightfully describe the relevant cell populations. Boxplots (n=294 for B cells, n=1,949 for monocytes and n=2,296 for T cells) show the median with interquartile range (IQR) (25–75%); whiskers extend 1.5x the IQR.
