Added new functions

DESCRIPTION

@@ -38,7 +38,8 @@ Imports:
     utils,
     gtools,
     stringr,
-    gridExtra
+    gridExtra,
+    ggcorrplot
 Suggests: 
     knitr,
     rmarkdown,

NAMESPACE

@@ -19,12 +19,14 @@ importFrom(edgeR,calcNormFactors)
 importFrom(edgeR,estimateDisp)
 importFrom(edgeR,glmFit)
 importFrom(edgeR,glmLRT)
+importFrom(ggcorrplot,ggcorrplot)
 importFrom(ggplot2,aes)
 importFrom(ggplot2,element_blank)
 importFrom(ggplot2,element_text)
 importFrom(ggplot2,facet_wrap)
 importFrom(ggplot2,geom_bar)
 importFrom(ggplot2,geom_boxplot)
+importFrom(ggplot2,geom_histogram)
 importFrom(ggplot2,geom_hline)
 importFrom(ggplot2,geom_jitter)
 importFrom(ggplot2,geom_point)
@@ -33,6 +35,7 @@ importFrom(ggplot2,ggsave)
 importFrom(ggplot2,ggtitle)
 importFrom(ggplot2,labs)
 importFrom(ggplot2,scale_colour_manual)
+importFrom(ggplot2,scale_fill_gradient2)
 importFrom(ggplot2,scale_shape_manual)
 importFrom(ggplot2,stat_summary)
 importFrom(ggplot2,theme)
@@ -45,11 +48,14 @@ importFrom(qs,qread)
 importFrom(reshape2,melt)
 importFrom(scales,percent)
 importFrom(stats,as.formula)
+importFrom(stats,complete.cases)
+importFrom(stats,cor)
 importFrom(stats,median)
 importFrom(stats,model.matrix)
 importFrom(stats,p.adjust)
 importFrom(stats,quantile)
+importFrom(stats,reshape)
 importFrom(stringr,str_sub)
 importFrom(utils,globalVariables)
 importFrom(viridis,scale_colour_viridis)
-importFrom(viridis,scale_fill_viridis)
+importFrom(viridis,scale_fill_viridis)

R/compute_downsampled_corr.r

@@ -0,0 +1,84 @@
+# Define global variables
+utils::globalVariables(c(".","dataset"))
+
+#' For a given down-sampled DE output, computes the correlation of the log-foldchange of the DEGs (at specified p-value) for a given dataset (celltype)
+
+#' @importFrom ggcorrplot ggcorrplot
+#' @importFrom data.table rbindlist setkey
+#' @importFrom stats reshape complete.cases cor
+#' @importFrom ggplot2 theme element_text scale_fill_gradient2 labs
+
+#' @param allstudies - a list containing DGE analysis outputs (subset to "celltype_all_genes" and the specified cell type) at each down-sampled point
+#' @param sampled - downsampling carried out based on what (either "individuals" or "cells")
+
+#' @return correlation matrix, plot and the number of DEGs at the specified p-value
+
+compute_downsampled_corr <- function(allstudies,
+                                     sampled="individuals"){
+
+    # check input parameters are fine
+    if(class(allstudies)!="list"){
+        stop("Error: allstudies should be a list.")
+    }
+
+    # variable to redefine allstudies
+    allstudies_new <- list()
+    j <- 1
+    # select data for each down-sampling point
+    for(sampling_point in names(allstudies)){
+        # redefine allstudies so each element only contains data for that down-sampled point
+        allstudies_new[[j]] <- allstudies[[sampling_point]]
+        names(allstudies_new)[[j]] <- sampling_point
+        j <- j+1
+    }
+    # reshape data so "dataset" is now a variable
+    allstudies_dt <- rbindlist(allstudies_new,idcol="dataset")
+    #filter dataset
+    setkey(allstudies_dt,name)
+
+    # make matrix for corr() - cols will be [datset, name, logFC]
+    mat_lfc <- allstudies_dt[,.(dataset,name,logFC)]
+    # reshape so cols now are [(gene) name, logFC.dataset1, logFC.dataset2,...,logFC.datasetN] (N being length(names(allstudies)))
+    mat_lfc <-
+    reshape(mat_lfc, idvar = "name", timevar = "dataset", 
+            direction = "wide")
+    # remove logFC. from name (now cols are just [name, dataset1, dataset2,...,datasetN] with logFC values in each column)
+    colnames(mat_lfc)<-
+        c(colnames(mat_lfc)[1],
+            substr(colnames(mat_lfc[,2:ncol(mat_lfc)]),
+                    7,nchar(colnames(mat_lfc[,2:ncol(mat_lfc)]))))
+    # remove NA rows
+    mat_lfc <- mat_lfc[complete.cases(mat_lfc), ]
+
+    # get correlation matrix
+    corr_lfc <- cor(mat_lfc[,2:ncol(mat_lfc)],method = "spearman")
+
+    # plot correlation matrix
+    if(sampled=="individuals"){
+        corr_plot.plot <- ggcorrplot(round(corr_lfc,2), 
+                hc.order = F, insig="pch",pch=5,pch.col = "grey",
+                pch.cex=9,
+                title=paste0("LogFC Correlation Matrix when downsampling individuals"),
+                colors = c("#FC4E07", "white", "#00AFBB"),
+                outline.color = "white", lab = TRUE, lab_size=3.5,
+                sig.level=0.05) + 
+                theme(plot.title = element_text(hjust = 0.7)) + 
+                scale_fill_gradient2(breaks=c(0, 1), limit=c(0, 1)) +
+                labs(fill="Corr")
+    }else{
+        corr_plot.plot <- ggcorrplot(round(corr_lfc,2), 
+        hc.order = F, insig="pch",pch=5,pch.col = "grey",
+        pch.cex=9,
+        title=paste0("LogFC Correlation Matrix when downsampling cells"),
+        colors = c("#FC4E07", "white", "#00AFBB"),
+        outline.color = "white", lab = TRUE, lab_size=3.5,
+        sig.level=0.05) + 
+        theme(plot.title = element_text(hjust = 0.7)) +
+        scale_fill_gradient2(breaks=c(0, 1), limit=c(0, 1)) +
+        labs(fill="Corr")
+    }
+
+    # output matrix, plot
+    return(list(corr_lfc, corr_plot.plot))
+
+}

R/downsampling_DEanalysis.r

@@ -23,7 +23,7 @@ utils::globalVariables(c("PValue","name"))
 #' @param pval_adjust_method the adjustment method for the p-value in the differential expression analysis. Default is benjamini hochberg "BH". See  stats::p.adjust for available options
 #' @param rmv_zero_count_genes whether genes with no count values in any cell should be removed. Default is TRUE
 
-#' @return saves all DE outputs for downsampled files as well as a summary table of results showing number of true DEGs detected at each number of samples/cells
+#' Saves all DE outputs for downsampled files as well as a summary table of results showing number of true DEGs detected at each number of samples/cells
 
 downsampling_DEanalysis <- function(data,
                                     range_downsampled="placeholder",