R/threshold.R

In dunbarlabNIH/barcodetrackR: Functions for Analyzing Cellular Barcoding Data

#' Threshold
#'
#' This is a helper which function takes in sequence data in table form, along with a threshold,
# " and returns new table with the criteria that each row contains at least one element
# " that registers above the threshold. Threshold here is defined as fractional contribution
#' to each column (e.g. if threshold is set as 0.0005, only rows in which an element is above 0.05% of
#' its column will be kept).
#'
#' @param your_data A data frame. Usually individual barcodes in rows and samples in columns.
#' @param thresh Numeric.
#' @param thresh_type Character. One of "relative" or "absolute"
#' @return A data frame where all rows (barcodes) that did not have at least one element meet the threshold have been discarded.
#' @examples
#' data(wu_subset)
#' threshold(SummarizedExperiment::assay(wu_subset, assay = "counts"),
#'     thresh = 0.0005
#' )
#' @export
threshold <- function(your_data, thresh = 0.0005, thresh_type = "relative") {
    if (thresh_type == "relative") {
        # makes vector of thresholds for each column
        threshes <- (colSums(your_data) * (thresh))

        # keeps rows that have at least one element larger than thresh for that column
        thresholded_data <- your_data[apply(your_data, 1, function(x) {
            any(threshes < x)
        }), ]
    }

    if (thresh_type == "absolute") {
        # makes vector of thresholds for each column
        threshes <- thresh

        # keeps rows that have at least one element larger than thresh for that column
        thresholded_data <- your_data[apply(your_data, 1, function(x) {
            any(threshes < x)
        }), ]
    }

    # in the unlikely case that a file has 0 reads, this will prevent the phantom
    # column from appearing in the final data frame
    thresholded_data[is.na(thresholded_data)] <- 0

    # returns subsetted data
    return(thresholded_data)
}