R/one2OneRnaMiRNA.R

In Mercedeh66/mirTarRnaSeq: mirTarRnaSeq

## Written by Mercedeh Movassagh <mercedeh@ds.dfci.harvard.edu>, Aug 2020

#' @importFrom dplyr filter
NULL

## quiet concerns of R CMD check regarding unbound global variables (in dplyr::filter() calls)
if (getRversion() >= "2.15.1") {
    utils::globalVariables(c("pvalue"))
}

## Inputs mRNA or miRNA dif expression file

#' one2OneRnaMiRNA correlation for miRNA and mRNA  using differntial expression
#' fold change and if/when available p-value
#'
#' This function inputs accept a list of dataframes and returns an obj with
#' two dataframes called FC and p-value. FC with rownames == genes and columns are FC1, 2, 3, ...
#' (with fold-changes) - P-value with rownames == genes and columns are P1, 2, 3, ... (with p-values)
#' both data.frames have the same order dimensions.
#' @param files a list of dataframes either miRNAs or mRNAs from various time points.
#' @param gene_colname Default is a vector character of length 1 "Gene" user can alter if they choose
#' This column contains the gene names.
#' @param fc_colname Default "FC" is coloumn name for fold changes user can alter if they choose.
#' @param pval_colname Default is "pvalue" column name for p-values (in input).
#' @param pthreshold P-value threshold.
#' @return Correlation dataframe
#' @export
#' @keywords correlation
#' @examples
#' \donttest{
#' x <- one2OneRnaMiRNA(files)
#' }
one2OneRnaMiRNA <- function(files, gene_colname = "Gene", fc_colname = "FC",
                            pval_colname = "pvalue", pthreshold = NULL) {


    # asser file......
    # assert is.character(gene_colname) && length(gene_colname) == 1
    # assert is.character(fc_colname) && length(fc_colname) == 1
    # assert (is.numeric(pthreshold) && length(pthreshold) == 1) || is.null(pthreshold)
    # assert ... gene col is.character! not factor.

    # do some mininal cleaning here...
    files <- lapply(files, function(x) {
        x <- as.data.frame(x)
        x[, gene_colname] <- as.character(x[, gene_colname]) # account for users...
        return(x)
    })

    # grab fold changes and gene names from each data.frame supplied
    foldchanges <- lapply(files, function(x) {
        ret <- x[, c(gene_colname, fc_colname)]
        rownames(ret) <- ret[, gene_colname, drop = TRUE]
        return(ret)
    })

    if (!is.null(pthreshold)) {
        # grab p-values and gene names from each data.frame supplied
        pvalues <- lapply(files, function(x) {
            ret <- x[, c(gene_colname, pval_colname)]
            rownames(ret) <- ret[, gene_colname, drop = TRUE]
            return(ret)
        })

        # produce set of genes significant in any of the gene lists
        # all genes concatenated in one long unique list
        siggenes <- unique(unlist(sapply(pvalues, function(x) {
            ret <- filter(x, pvalue < pthreshold)[, gene_colname, drop = TRUE]
            return(ret)
        }))) # for some reason sapply doesn't simplify here... hence the 'unlist'
    } else {
        # get set of genes regardless of p-value
        pvalues <- NULL
        siggenes <- unique(unlist(sapply(foldchanges, function(x) {
            x[, gene_colname, drop = TRUE] # we're producing a long vector here...
        })))
    }

    # filter siggenes for gene names present in all input lists!
    for (x in foldchanges) {
        siggenes <- intersect(siggenes, x[, gene_colname, drop = TRUE]) # we're intersecting _vectors_ here
    }

    # filter and reorder FCs by sig genes
    foldchanges <- lapply(foldchanges, function(x) {
        return(x[siggenes, fc_colname, drop = FALSE])
    })

    # make one data frame with _just_ the FC columns.
    foldchanges <- do.call(cbind.data.frame, foldchanges)
    colnames(foldchanges) <- paste("FC", seq_len(ncol(foldchanges)), sep = "") # cleanup column names

    if (!is.null(pthreshold)) {
        # filter and reorder pvalues by sig genes
        pvalues <- lapply(pvalues, function(x) {
            return(x[siggenes, pval_colname, drop = FALSE])
        })

        # make one data frame with _just_ the pvalue columns.
        pvalues <- do.call(cbind.data.frame, pvalues)
        colnames(pvalues) <- paste("P", seq_len(ncol(pvalues)), sep = "") # cleanup column names
    }

    # now 'foldchanges' and 'pvalues' are _data.frame_ with _just_ FC/P columns.
    return(list(foldchanges = foldchanges, pvalues = pvalues))
}