R/plots.R: change in make_html_pca and make_pca and create_plots to handle more than 1 factor

R/plots.R

@@ -4,18 +4,47 @@
 #' @param percent_var Vector containing the percentage of variance
 #' of PC1 and PC2
 #' @param output_folder Folder where the figure will be created
+#' @param columns The design columns
 #' @import plotly
 #' @import htmlwidgets
-make_html_pca <- function(pca_data, percent_var, output_folder) {
-    n <- length(unique(pca_data$condition))
+make_html_pca <- function(pca_data, percent_var, output_folder, columns) {
+    n <- length(unique(pca_data[[columns[1]]]))
     hues <- seq(15, 375, length = n + 1)
     color_list <- hcl(h = hues, l = 65, c = 100)[1:n]
+    hover <- NULL
+    for (c in columns) {
+        hover <- paste0(hover, "<br>", c, " : ", pca_data[[c]])
+    }
+    pca_data$hover <- hover
+    if (length(columns) == 1) {
+        ppca <- plot_ly(
+            data = pca_data, x = ~PC1, y = ~PC2,
+            color = as.formula(paste0("~", columns[1])),
+            colors = color_list, type = "scatter",
+            text = ~hover, mode = "markers",
+            marker = list(size = 30)
+        )
+    } else if (length(columns) == 2) {
         ppca <- plot_ly(
             data = pca_data, x = ~PC1, y = ~PC2,
-        color = ~condition, colors = color_list, type = "scatter",
-        text = ~name, mode = "markers",
+            color = as.formula(paste0("~", columns[1])),
+            symbol = as.formula(paste0("~", columns[2])),
+            colors = color_list, type = "scatter",
+            text = ~hover, mode = "markers",
             marker = list(size = 30)
-    ) %>%
+        )
+    } else {
+        pca_data[["tmp"]] <- (as.numeric(pca_data[[columns[3]]]) * 7)
+        ppca <- plot_ly(
+            data = pca_data, x = ~PC1, y = ~PC2,
+            color = as.formula(paste0("~", columns[1])),
+            symbol = as.formula(paste0("~", columns[2])),
+            marker = list(size = pca_data$tmp),
+            colors = color_list, type = "scatter",
+            text = ~hover, mode = "markers"
+        )
+    }
+    ppca <- ppca %>%
         layout(
             title = paste0("PCA plot"),
             margin = list(l = 100, r = 100, b = 100, t = 100, pad = 4),
@@ -44,6 +73,59 @@ make_html_pca <- function(pca_data, percent_var, output_folder) {
 }
 
 
+#' Make ggplot PCA
+#'
+#' @param pca_data Object returned by DESeq2::plotPCA containing PCA table
+#' @param percent_var Vector containing the percentage of variance
+#' of PC1 and PC2
+#' @param columns A vector containing the columns to display in PCA data
+#' @param dds a DESeqDataSet object
+#' @import ggplot2
+make_pca <- function(pca_data, percent_var, columns, dds) {
+    if (length(columns) == 1) {
+        pggpca <- ggplot2::ggplot(pca_data, aes_string("PC1", "PC2", color = columns[1]))
+    } else if (length(columns) == 2) {
+        pggpca <- ggplot2::ggplot(
+            pca_data,
+            aes_string("PC1", "PC2", color = columns[1], shape = columns[2])
+        )
+    } else {
+        list_plots <- list()
+        my_combinations <- combn(columns, 2)
+        counter <- 1
+        for (c in seq_len(ncol(my_combinations))) {
+            cond1 <- my_combinations[1, c]
+            cond2 <- my_combinations[2, c]
+            i <- which(columns == cond1)
+            n <- length(unique(dds[[columns[i]]]))
+            hues <- seq(round(15 + 70 * (i - 1)), round(375 + 150 * (i - 1)),
+                length = n + 1
+            )
+            tmpcols <- hcl(h = hues, l = 65 - (i - 1) * 10, c = 100)[1:n]
+            tmp <- ggplot2::ggplot(
+                pca_data,
+                aes_string("PC1", "PC2", color = cond1, shape = cond2)
+            ) +
+                scale_color_manual(values = tmpcols) +
+                geom_point(size = 3) +
+                xlab(paste0("PC1: ", percent_var[1], "% variance")) +
+                ylab(paste0("PC2: ", percent_var[2], "% variance")) +
+                coord_fixed()
+            list_plots[[counter]] <- tmp
+            counter <- counter + 1
+        }
+        return(list_plots)
+    }
+    if (length(columns) == 1 || length(columns) == 2) {
+        pggpca <- pggpca +
+            geom_point(size = 3) +
+            xlab(paste0("PC1: ", percent_var[1], "% variance")) +
+            ylab(paste0("PC2: ", percent_var[2], "% variance")) +
+            coord_fixed()
+        return(list(pggpca))
+    }
+}
+
 
 #' Plots of the experiment
 #'
@@ -55,17 +137,21 @@ make_html_pca <- function(pca_data, percent_var, output_folder) {
 #' @import plotly
 #' @import htmlwidgets
 #' @import RColorBrewer
+#' @import stringr
 #' @export
 create_plots <- function(dds, output_folder) {
+    my_formula <- paste(dds@design, collapse = "")
+    columns <- unique(str_split(str_replace_all(my_formula, "[~\\s]", ""), 
+                      "([\\+\\*\\:\\-\\^]|%in%)")[[1]])
     rld <- DESeq2::rlogTransformation(dds, blind = TRUE)
-    n <- length(unique(dds$condition))
+    n <- length(unique(dds[[columns[1]]]))
     hues <- seq(15, 375, length = n + 1)
     condcols <- hcl(h = hues, l = 65, c = 100)[1:n]
     counts_normalise <- DESeq2::counts(dds, normalized = T)
-    names(condcols) <- unique(dds$condition)
+    names(condcols) <- unique(dds[[columns[1]]])
 
     pca_data <- DESeq2::plotPCA(rld,
-        intgroup = c("condition"),
+        intgroup = columns,
         returnData = TRUE
     )
     percent_var <- round(100 * attr(pca_data, "percentVar"))
@@ -75,37 +161,38 @@ create_plots <- function(dds, output_folder) {
     colours <- colorRampPalette(rev(RColorBrewer::brewer.pal(9, "Blues")))(255)
 
     cts_norm_df <- as.data.frame(counts_normalise)
-    annotation_col <- data.frame(condition = dds$condition)
-    rownames(annotation_col) <- colnames(dds)
-    mat_colors <- list(
-        condition = condcols
+    annotation_col <- data.frame("row_names" = colnames(dds))
+    mat_colors <- NULL
+    for (i in seq_len(length(columns))) {
+        n <- length(unique(dds[[columns[i]]]))
+        hues <- seq(round(15 + 70 * (i - 1)), round(375 + 150 * (i - 1)),
+            length = n + 1
         )
-    names(mat_colors$condition) <- sort(unique(dds$condition))
+        tmpcols <- hcl(h = hues, l = 65 - (i - 1) * 10, c = 100)[1:n]
+        mat_colors[[columns[i]]] <- tmpcols
+        names(mat_colors[[columns[i]]]) <- sort(unique(dds[[columns[i]]]))
+        annotation_col[columns[i]] <- dds[[columns[i]]]
+    }
+    rownames(annotation_col) <- colnames(dds)
+    annotation_col$row_names <- NULL
 
     pdf(paste0(output_folder, "/plots_general_view.pdf"))
 
     barplot(colSums(DESeq2::counts(dds, normalized = F)),
-        col = condcols[as.factor(dds$condition)],
+        col = condcols[as.factor(dds[[columns[1]]])],
         las = 2, cex.names = 0.4,
         main = "Pre Normalised Counts"
     )
     barplot(colSums(DESeq2::counts(dds, normalized = T)),
-        col = condcols[as.factor(dds$condition)],
+        col = condcols[as.factor(dds[[columns[1]]])],
         las = 2, cex.names = 0.4,
         main = "Post Normalised Counts"
     )
 
-    pggpca <- ggplot2::ggplot(pca_data, aes(PC1, PC2, color = condition)) +
-        geom_point(size = 3) +
-        xlab(paste0("PC1: ", percent_var[1], "% variance")) +
-        ylab(paste0("PC2: ", percent_var[2], "% variance")) +
-        coord_fixed() +
-        geom_label_repel(
-            data = pca_data, alpha = 0.7,
-            mapping = aes(PC1, PC2, label = name), size = 1.2,
-        )
-    plot(pggpca)
-
+    pca_plots <- make_pca(pca_data, percent_var, columns, dds)
+    for (p_pca in pca_plots) {
+        print(p_pca)
+    }
 
     gplots::heatmap.2(sample_dist_matrix,
         trace = "none", col = colours,
@@ -119,9 +206,8 @@ create_plots <- function(dds, output_folder) {
         annotation_colors = mat_colors, angle_col = "45"
     )
 
-
     DESeq2::plotDispEsts(dds)
 
     dev.off()
-    make_html_pca(pca_data, percent_var, output_folder)
+    make_html_pca(pca_data, percent_var, output_folder, columns)
 }