R/methods-cellTypesPerCluster.R

In roryk/bcbioSinglecell: Single-Cell RNA-Seq Utilities

#' Cell Types per Cluster
#'
#' @name cellTypesPerCluster
#' @family Clustering Functions
#' @author Michael Steinbaugh
#'
#' @inheritParams general
#'
#' @param min Minimum number of marker genes per cluster.
#' @param max Maximum number of marker genes per cluster.
#'
#' @return `grouped_df` grouped by "`cluster`", containing the count (`n`) of
#'   significant known makers per cell type.
#'
#' @examples
#' # grouped_df ====
#' x <- cellTypesPerCluster(known_markers_small)
#' glimpse(x)
NULL



# Methods ======================================================================
#' @rdname cellTypesPerCluster
#' @export
setMethod(
    "cellTypesPerCluster",
    signature("grouped_df"),
    function(
        object,
        min = 1L,
        max = Inf
    ) {
        assert_are_identical(attr(object, "vars"), "cellType")
        requiredCols <- c(
            "cellType",  # bcbio
            "cluster",   # Seurat
            "geneID",    # bcbio
            "geneName",  # bcbio
            "avgLogFC",  # Seurat v2.1
            "padj"       # Seurat v2.1
        )
        assert_is_subset(requiredCols, colnames(object))

        # Note that the order is important here
        groupCols <- c("cluster", "cellType")

        data <- object %>%
            ungroup() %>%
            select(!!!syms(groupCols), everything()) %>%
            mutate_at(groupCols, as.factor) %>%
            group_by(!!!syms(groupCols)) %>%
            arrange(!!sym("padj"), .by_group = TRUE) %>%
            # Use `toString()` instead of `aggregate()` for R Markdown tables
            summarize(
                n = n(),
                # Genes are arranged by P value
                geneID = toString(!!sym("geneID")),
                geneName = toString(!!sym("geneName"))
            ) %>%
            group_by(!!sym("cluster")) %>%
            arrange(desc(!!sym("n")), .by_group = TRUE)

        # Apply minimum and maximum gene cutoffs
        if (is.numeric(min) && min > 1L) {
            data <- filter(data, !!sym("n") >= !!min)
        }
        if (is.numeric(max) && max > 1L) {
            data <- filter(data, !!sym("n") <= !!max)
        }
        assert_has_rows(data)

        data
    }
)