R/jj_plot_fc_fc.R

In mathosi/jj: Just Joise

#' show results from differential expression analysis
#'
#' jj_plot_fc_fc plot fold changes from two different DEG analyses against each other
#'
#' @name jj_plot_fc_fc
#' @param logfc_column sparse matrix, vector, or data.frame
#' @param pval_column vector with group annotation, has to have length equal to ncol(sparse_mat)/nrow(data.frame)/length(vector)
#' @param symbol_column provide `symbol_column` to add text to all points that are highlighted (when use_text=TRUE)
#' @param labs_range vector specifying range of x axis (and for fc_fc plot range of y axis)
#' @param marker_thres define `marker_thres` to highlight all points with absolute value above
#' @param col_vec define larger, smaller, intermediate and highlight colours
#' @param markers_highlight vector of genes from the symbol_column which should additionally be highlighted
#' @param markers_highlight_col named vector assigning a colour to each gene to be highlighed
#' @param only_highlight if TRUE, only show markers specified in `markers_highlight`
#' @param col_by_highlight if TRUE, only colour the markers specified in `markers_highlight`
#' @param use_text plot text if there are any points to highlight
#' @param add_thres_line if TRUE, plot dashed lines at the defined `marker_thres`
#' @param alpha alpha level for points (working?)
#' @returns ggplot scatterplot of log10(pval) vs FC or FC vs FC
#' @export
#' @examples
#' #Using Seurats FindMarkers()
#' library(Seurat)
#' marker_df = FindMarkers(pbmc_small, group.by = 'groups', ident.1 = 'g1', ident.2 = 'g2')
#' marker_df$symbol = rownames(marker_df)
#' marker_df2 = FindMarkers(pbmc_small, group.by = 'RNA_snn_res.1', ident.1 = '0', ident.2 = '1')
#' marker_df2$symbol = rownames(marker_df2)
#' markers_joined = dplyr::inner_join(marker_df, marker_df2, by = 'symbol')
#' #or using a random data.frame
#' markers_joined = data.frame(avg_log2FC.x = rnorm(100, 0, 2), avg_log2FC.y = rnorm(100, 0, 2), symbol = paste0(sample(LETTERS, 100, replace = T), sample(letters, 100, replace = T), 1:100))
#'# simple logFC-logFC plot with highlighting above |avg_log2FC| = 1
#'jj_plot_fc_fc(plot_df = markers_joined, logfc_column1 = 'avg_log2FC.x', logfc_column2 = 'avg_log2FC.y', marker_thres = 1)    
#'# adding range to plot and text to markers, also adding threshold lines
#'jj_plot_fc_fc(markers_joined, logfc_column1 = 'avg_log2FC.x', logfc_column2 = 'avg_log2FC.y',symbol_column = 'symbol', labs_range = c(-4,4, -4,4), marker_thres = 2, use_text = T, add_thres_line = T) 
#' #specifying custom highlight
#'jj_plot_fc_fc(markers_joined, logfc_column1 = 'avg_log2FC.x', logfc_column2 = 'avg_log2FC.y',symbol_column = 'symbol', labs_range = c(-4,4, -4,4), marker_thres = 2, use_text = T,  markers_highlight = markers_joined$symbol[1:4], col_by_highlight = T) 
#' #changing colour and size of custom highlight
#'jj_plot_fc_fc(markers_joined, logfc_column1 = 'avg_log2FC.x', logfc_column2 = 'avg_log2FC.y',symbol_column = 'symbol',
#'              labs_range = c(-4,4, -4,4), marker_thres = Inf, use_text = T, markers_highlight = markers_joined$symbol[1:4], 
#'              col_by_highlight = T,highlight.pt.size = 3, markers_highlight_col = structure(rep('orange',4), names=markers_joined$symbol[1:4])) + guides(colour = 'none')


#' @export 
jj_plot_fc_fc =function(plot_df, logfc_column1, logfc_column2, symbol_column=NULL, 
                        labs_range = c(-2, 2, -2, 2), marker_thres = 0.5, 
                        col_vec = c("red", "blue", "black","green"),
                        alpha=1, pt.size=1, highlight.pt.size=1, only_highlight=FALSE,
                        markers_highlight=NULL, col_by_highlight=FALSE, use_text=FALSE,
                        add_thres_line = FALSE, markers_highlight_col=NULL){
  #stopifnot(colour_aes %in% colnames(plot_df))
  library(ggrepel)
  symbols_use = c('outside plotting area', 'inside plotting area')
  
  plot_df = plot_df[, c(logfc_column1, logfc_column2, symbol_column)] #ensure no unnamed columns are in the df -> results in attempt to use zero-length variable name
  
  if(!is.null(labs_range)){
    if(length(labs_range) == 2){
      labs_range = c(labs_range, labs_range)
    }
    #add a column that is used for different symbols (circle if inside plot, triangle if outside)
    plot_df =  dplyr::mutate(plot_df, localization = ifelse(!!sym(logfc_column1) >= labs_range[2] | !!sym(logfc_column2) >= labs_range[4], symbols_use[1],
                                                            ifelse(!!sym(logfc_column1) <= labs_range[1] | !!sym(logfc_column2) <= labs_range[3],
                                                                   symbols_use[1], symbols_use[2])))
    plot_df$localization = factor(plot_df$localization, levels = c('inside plotting area', 'outside plotting area'))
    
    #cap the values outside of plot
    plot_df[, logfc_column1] = ifelse(plot_df[, logfc_column1] >= labs_range[2], labs_range[2],
                                      ifelse(plot_df[, logfc_column1] <= labs_range[1],
                                             labs_range[1], plot_df[, logfc_column1]))
    plot_df[, logfc_column2] = ifelse(plot_df[, logfc_column2] >= labs_range[4], labs_range[4],
                                      ifelse(plot_df[, logfc_column2] <= labs_range[3],
                                             labs_range[3], plot_df[, logfc_column2]))
  }
  
  #give label for all genes > marker_thres
  #give label for all genes > marker_thres
  mthres = as.character(marker_thres)
  small_val = paste0(">= ", mthres)
  middle_val = sprintf(">-%s & <%s", mthres, mthres)
  big_val = paste0("<= -", mthres)
  
  plot_df$logFC = ifelse(plot_df[, logfc_column1] >= marker_thres & plot_df[, logfc_column2] >= marker_thres, big_val,
                         ifelse(plot_df[, logfc_column1]<=-marker_thres & plot_df[, logfc_column2]<=-marker_thres ,
                                small_val, middle_val))
  
  
  if(!is.null(symbol_column)){
    plot_df$label_use <- ''
    plot_df$label_use[plot_df$logFC != middle_val] <- plot_df[, symbol_column][plot_df$logFC != middle_val]
    
    if(!is.null(markers_highlight) & any(plot_df[, symbol_column] %in% markers_highlight)){
      if(only_highlight){
        plot_df = plot_df[plot_df[, symbol_column] %in% markers_highlight, ]
      }
      #markers_highlight = unique(plot_df[, symbol_column])
      plot_df$label_use[plot_df[, symbol_column] %in% markers_highlight] <- plot_df[, symbol_column][plot_df[, symbol_column] %in% markers_highlight]
      plot_df$logFC[plot_df[, symbol_column] %in% markers_highlight] = 'highlight'
      plot_df = plot_df[order(plot_df[, symbol_column] %in% markers_highlight, decreasing = F), ]
    }
  }
  
  cols_use = structure(col_vec, names = c(big_val, small_val, middle_val, "highlight"))
  cols_use = cols_use[names(cols_use) %in% plot_df$logFC]
  
  if(col_by_highlight){
    markers_highlight = markers_highlight[markers_highlight %in% plot_df[, symbol_column]]
    if(is.null(markers_highlight_col)){
      cols_use = jj_get_jj_colours(markers_highlight)
    }else{
      cols_use = markers_highlight_col[names(markers_highlight_col) %in% markers_highlight]
    }
    
    if(!is.null(labs_range)){
      g1 = ggplot() +
        geom_point(data=plot_df, aes_string(x=logfc_column1, y=logfc_column2, shape='localization'),alpha=alpha, size=pt.size) +
        geom_point(data=plot_df[plot_df[, symbol_column] %in% markers_highlight, ], aes_string(x=logfc_column1, y=logfc_column2,
                                                                                               shape='localization', 
                                                                                               colour='label_use'), size=highlight.pt.size)
    }else{
      g1 =  ggplot() +
        geom_point(data=plot_df, aes_string(x=logfc_column1, y=logfc_column2), alpha=alpha, size=pt.size) +
        geom_point(data=plot_df[plot_df[, symbol_column] %in% markers_highlight, ], aes_string(x=logfc_column1, y=logfc_column2,
                                                                                               colour='label_use'), size=highlight.pt.size)
    }
  }else{
    if(!is.null(labs_range)){
      g1 = ggplot(plot_df, aes_string(x=logfc_column1, y=logfc_column2)) +
        geom_point(aes(colour = logFC, shape=localization), size=pt.size, alpha=alpha)
    }else{
      g1 = ggplot(plot_df, aes_string(x=logfc_column1, y=logfc_column2), alpha=alpha) +
        geom_point(aes(colour = logFC), size=pt.size)
    }
  }
  
  if(!is.null(symbol_column) & use_text){
    g1 = g1 + ggrepel::geom_text_repel(data=plot_df,  aes_string(x=logfc_column1, y=logfc_column2, label='label_use'), max.overlaps = 500)
  }
  
  g1 = g1 +
    scale_colour_manual(values = cols_use) +
    geom_hline(yintercept = 0) + geom_vline(xintercept = 0) + theme_minimal()
  
  if(!is.null(labs_range)){
    g1 = g1 + coord_fixed(xlim=labs_range[1:2], ylim = labs_range[3:4], expand = F, clip = 'off')
  }
  
  if(add_thres_line){
    g1 = g1 + 
      geom_vline(xintercept = -marker_thres, linetype="dashed", color ='black', size = 0.5) +
      geom_vline(xintercept = marker_thres, linetype="dashed", color ='black', size = 0.5) +
      geom_hline(yintercept = -marker_thres, linetype="dashed", color ='black', size = 0.5) +
      geom_hline(yintercept = marker_thres, linetype="dashed", color ='black', size = 0.5)
  }
  
  return(g1)
}