R/densityPlot.R

In cbiagii/pcitRif: Coexpression for Transcription Factors using Regulatory Impact Factors and Partial Correlation and Information Theory analysis

#' @title Density distribution of correlation coefficients and significant PCIT values
#'
#' @description Generate the density plot for adjacency matrices. This
#' function uses the raw adjacency matrix and significant adjacency
#' matrix resulted from \code{\link{PCIT}} function.
#'
#' @param mat1 Raw adjacency matrix.
#' @param mat2 Significant adjacency matrix.
#' @param threshold Threshold of correlation module to plot (default: 0.5).
#'
#' @return Returns an density plot of raw correlation with significant PCIT values.
#'
#' @importFrom ggplot2 geom_histogram scale_fill_manual xlab ylab element_line element_blank geom_density geom_vline aes_string
#' @importFrom ggpubr ggarrange
#' @importFrom stats density
#' 
#' @examples
#' # loading a simulated normalized data
#' data('simNorm')
#'
#' # getting the PCIT results
#' results <- PCIT(simNorm[1:20, ])
#'
#' # using the PCIT results to get density distribution of correlation coefficients
#' densityPlot(mat1 = results$adj_raw,
#'             mat2 = results$adj_sig,
#'             threshold = 0.5)
#'
#' @export
densityPlot <- function(mat1, mat2, threshold = 0.5) {
    if (!is.data.frame(mat1) & !is.matrix(mat1)) {
        stop("mat1 must be a dataframe or a matrix")
    }
    if (!is.data.frame(mat2) & !is.matrix(mat2)) {
        stop("mat2 must be a dataframe or a matrix")
    }
    
    df1 <- data.frame(corr = mat1[upper.tri(mat1)])
    df2 <- data.frame(corr = mat1[upper.tri(mat1) & !is.na(mat1)], sig = "All")
    df3 <- data.frame(corr = mat2[intersect(which(as.single(mat2) != 0), 
        which(upper.tri(mat1)))], sig = "PCIT Significant")
    df4 <- data.frame(corr = mat1[intersect(which(abs(mat1) > threshold), 
        which(upper.tri(mat1)))], sig = paste0("abs. cor. > ", threshold))
    
    dn <- density(df1[["corr"]])
    pt1 <- ggplot(df1, aes(x = .data[["corr"]])) + geom_density(colour = "black", 
        fill = "#a0b8d6", size = 1) + scale_x_continuous(name = "Correlation Coefficient", 
        breaks = seq(-1, 1, 0.2), limits = c(-1, 1)) + scale_y_continuous(name = "Density", 
        expand = c(0, 0), limits = c(0, range(dn[["y"]] + 0.1)[2])) + ggtitle("Density Plot of Raw Correlation Coefficients") + 
        theme_bw() + theme(axis.line = element_line(size = 1, colour = "black"), 
        panel.grid.major = element_line(colour = "#d3d3d3"), panel.grid.minor = element_blank(), 
        panel.border = element_blank(), panel.background = element_blank(), 
        plot.title = element_text(size = 14, face = "bold"), axis.text.x = element_text(colour = "black", 
            size = 9), axis.text.y = element_text(colour = "black", size = 9)) + 
        geom_vline(xintercept = mean(df1[["corr"]]), size = 1, colour = "#FF3721", 
            linetype = "dashed")
    
    df <- rbind(df2, df3)
    pt2 <- ggplot(df, aes_string("corr")) + geom_histogram(data = subset(df, 
        df[["sig"]] == "All"), aes(fill = subset(df, df[["sig"]] == "All")[["sig"]]), 
        breaks = seq(-1, 1, by = 0.05), col = "black") + geom_histogram(data = subset(df, 
        df[["sig"]] == "PCIT Significant"), aes(fill = subset(df, df[["sig"]] == 
        "PCIT Significant")[["sig"]]), breaks = seq(-1, 1, by = 0.05), 
        col = "black") + scale_fill_manual(name = "", values = c("gray", 
        "black"), labels = c("All", "PCIT Significant")) + ggtitle("Density Distribution of Correlation Coefficients") + 
        xlab("Correlation Coefficient") + ylab("Frequency") + scale_y_continuous(expand = c(0, 
        0)) + theme_bw() + theme(axis.line = element_line(size = 1, colour = "black"), 
        panel.grid.major = element_line(colour = "#d3d3d3"), panel.grid.minor = element_blank(), 
        panel.border = element_blank(), panel.background = element_blank(), 
        plot.title = element_text(size = 14, face = "bold"), axis.text.x = element_text(colour = "black", 
            size = 9), axis.text.y = element_text(colour = "black", size = 9), 
        legend.position = "top")
    
    df <- rbind(df2, df3, df4)
    pt3 <- ggplot(df, aes_string("corr")) + geom_histogram(data = subset(df, 
        df[["sig"]] == "All"), aes(fill = subset(df, df[["sig"]] == "All")[["sig"]]), 
        breaks = seq(-1, 1, by = 0.05), col = "black") + geom_histogram(data = subset(df, 
        df[["sig"]] == paste0("abs. cor. > ", threshold)), aes(fill = subset(df, 
        df[["sig"]] == paste0("abs. cor. > ", threshold))[["sig"]]), breaks = seq(-1, 
        1, by = 0.05), col = "black") + geom_histogram(data = subset(df, 
        df[["sig"]] == "PCIT Significant"), aes(fill = subset(df, df[["sig"]] == 
        "PCIT Significant")[["sig"]]), breaks = seq(-1, 1, by = 0.05), 
        col = "black") + scale_fill_manual(name = "", values = c("red", 
        "gray", "black"), labels = c(paste0("abs. cor. > ", threshold), 
        "All", "PCIT Significant")) + xlab("Correlation Coefficient") + 
        ylab("Frequency") + scale_y_continuous(expand = c(0, 0)) + theme_bw() + 
        theme(axis.line = element_line(size = 1, colour = "black"), panel.grid.major = element_line(colour = "#d3d3d3"), 
            panel.grid.minor = element_blank(), panel.border = element_blank(), 
            panel.background = element_blank(), plot.title = element_text(size = 14, 
                face = "bold"), axis.text.x = element_text(colour = "black", 
                size = 9), axis.text.y = element_text(colour = "black", 
                size = 9), legend.position = "top")
    
    df <- rbind(df2, df3, df4)
    pt4 <- ggplot(df, aes_string("corr")) + geom_histogram(data = subset(df, 
        df[["sig"]] == "All"), aes(fill = subset(df, df[["sig"]] == "All")[["sig"]]), 
        breaks = seq(-1, 1, by = 0.05), col = "black", alpha = 0.8) + geom_histogram(data = subset(df, 
        df[["sig"]] == "PCIT Significant"), aes(fill = subset(df, df[["sig"]] == 
        "PCIT Significant")[["sig"]]), breaks = seq(-1, 1, by = 0.05), 
        col = "black", alpha = 0.9) + geom_histogram(data = subset(df, 
        df[["sig"]] == paste0("abs. cor. > ", threshold)), aes(fill = subset(df, 
        df[["sig"]] == paste0("abs. cor. > ", threshold))[["sig"]]), breaks = seq(-1, 
        1, by = 0.05), col = "black", alpha = 0.8) + scale_fill_manual(name = "", 
        values = c("gray10", "gray", "darkred"), labels = c(paste0("abs. cor. > ", 
            threshold), "All", "PCIT Significant")) + xlab("Correlation Coefficient") + 
        ylab("Frequency") + scale_y_continuous(expand = c(0, 0)) + theme_bw() + 
        theme(axis.line = element_line(size = 1, colour = "black"), panel.grid.major = element_line(colour = "#d3d3d3"), 
            panel.grid.minor = element_blank(), panel.border = element_blank(), 
            panel.background = element_blank(), plot.title = element_text(size = 14, 
                face = "bold"), axis.text.x = element_text(colour = "black", 
                size = 9), axis.text.y = element_text(colour = "black", 
                size = 9), legend.position = "top")
    
    pt <- ggarrange(plotlist = list(pt1, pt2, pt3, pt4), widths = c(2, 
        2))
    
    return(pt)
}