Multicore solution for Seurat FindAllMarkers()

mcFindMarkers.R

n_clust <- 1:(max(as.numeric(Idents(seurat_obj))))
mcFindMarkers <- function(i){
  ident1 <- i
  ident2 <- n_clust[n_clust != i]
  table <- FindMarkers(seurat_obj,
    ident.1 = ident1, ident.2 = ident2, only.pos = TRUE)
  table$Gene.name.uniq <- rownames(table)
  table$cluster <- rep(i, nrow(table))
  return(table)
}

marker_results <- list()[n_clust]
ptm <- proc.time()
marker_results <- parallel::mclapply(n_clust, mcFindMarkers, mc.cores = 16)
time_diff <- proc.time() - ptm
time_diff

# nice way to flatten list into a single DF
markers <- dplyr::bind_rows(markers)

Hi, just a minor comment: in my case Idents(seurat_obj) is a factor, with levels from 0 to 19, so instead of using 1:(max(as.numeric(Idents(seurat_obj)))), I used Idents(seurat_obj) %>% unique() %>% as.character() %>% as.numeric(), otherwise the factor to numeric conversion does not give what I need.

Very helpful--thank you! BTW, I think that markers <- dplyr::bind_rows(markers) should be markers <- dplyr::bind_rows(marker_results)