R/regress_features.R

In whtns/seuratTools: Tools for Managing Seurat Objects

#' Regress Seurat Object by Given Set of Genes
#'
#' @param seu A seurat object
#' @param feature_set
#' @param set_name as a string
#' @param regress
#'
#' @return
#' @export
#'
#' @examples
#'
#' regressed_seu <- regress_by_features(panc8, feature_set = cc.genes$s.genes, set_name = "s_genes")
#'
regress_by_features <- function(seu, feature_set, set_name, regress = TRUE, ...) {
    message(paste0("regressing seurat objects by ", set_name))

    if (!is.list(feature_set)) feature_set <- list(feature_set)

    ctrl <- 100
    if (dim(seu)[2] < 100) {
        ctrl <- dim(seu)[2] / 10
    }

    seu <- AddModuleScore(seu, feature_set, name = set_name, ctrl = ctrl)

    set_name <- paste0(set_name, length(feature_set))
    message(paste0("Module score stored as ", set_name))

    if ("integrated" %in% names(seu@assays)) {
        default_assay <- "integrated"
    } else {
        default_assay <- "gene"
    }

    # revert default assay to "integrated"
    Seurat::DefaultAssay(seu) <- default_assay

    # regress out feature set
    if (regress) {
        seu <- ScaleData(seu, vars.to.regress = set_name)
        # rerun dimensional reduction
        reductions <- names(seu@reductions)
        resolutions <- stringr::str_extract(names(seu[[]])[grepl("snn", names(seu[[]]))], "[0-9].*$")
        resolutions <- sort(as.numeric(resolutions))

        seu <- seurat_reduce_dimensions(seu)
        seu <- seurat_cluster(seu, resolution = resolutions)
    }

    return(seu)
}