Presto performs a fast Wilcoxon rank sum test and auROC analysis. Latest benchmark ran 1 million observations, 1K features, and 10 groups in 16 seconds (sparse input) and 85 seconds (dense input).
the logfc was calculated same as seurat now.
We are working on getting presto into CRAN. For now, install Presto from github directly:
library(devtools)
install_github('microbiomeCat/presto')
Run presto on a matrix, Seurat, or SingleCellExperiment input object.
wilcoxauc(X, y)
wilcoxauc(seurat_object, 'group_name')
wilcoxauc(sce_object, 'group_name')
For examples, see ?wilcoxauc and the vignette
markers <- wilcoxauc(seurat_object, 'seurat_clusters',seurat_assay="RNA",assay="data") %>%
filter(logFC >= 0.25) %>% filter(pct_in >= 10) %>% filter(pval <= 0.01)
markers$padj <- p.adjust(markers$pval, method='bonferroni', n=nrow(seurat_object@assays$RNA@data))
markers <- markers[,c("pval","logFC","pct_in","pct_out","padj","group","feature")]
colnames(markers) <- c("p_val","avg_log2FC","pct.1","pct.2","p_val_adj","cluster","gene")
markers$pct.1 <- round(markers$pct.1/100,3)
markers$pct.2 <- round(markers$pct.2/100,3)