Merge pull request #7 from FridleyLab/oospina_dev

multiple fixes

R/STList.R

@@ -811,7 +811,7 @@ process_lists = function(counts_df_list, coords_df_list){
     # If no counts in the entire FOV, mark for removal
     if(sum(coords_df_list[[name_i]][['total_counts']]) < 1){
       rm_fov = append(rm_fov, name_i)
-      warning(paste0('No counts present in FOV ', name_i, '. Removong from data set.\n'))
+      warning(paste0('No counts present in FOV ', name_i, '. Removing from data set.\n'))
     }
   }
 

R/distribution_plots.R

@@ -1,23 +1,25 @@
 ##
 #' @title per_unit_counts: Generates distribution plots of spot/cell meta data or gene expression
-#' @description Generates violin plots, boxplots, or density plots of continuous
-#' variables in the spatial meta data or of gene expression to make comparisons
-#' across samples
+#' @description Generates violin plots, boxplots, or density plots of variables in the
+#' spatial meta data or of gene expression
 #' @details The function allows to visualize the distribution of spot/cell total
 #' counts, total genes, or expression of specific genes across all samples for
-#' comparative purposes.
+#' comparative purposes. It also allows grouping of gene expression values by
+#' categorical variables (e.g., clusters).
 #'
 #' @param x an STlist
 #' @param plot_meta vector of variables in `x@spatial_meta` to plot distributions.
 #' If 'total_counts', the function plots the counts per spot/cell. If 'total_genes',
-#' the function plots the number of genes per spot/cell are plotted.
-#' @param genes vector of genes to plot expression distribution
+#' the function plots the number of genes per spot/cell are plotted
+#' @param genes vector of genes to plot expression distribution. If used in conjunction
+#' with `plot_meta`, the expression values are grouped using that variable
 #' @param samples samples to include in the plot. Default (NULL) includes all samples
 #' @param data_type one of 'tr' or 'raw', to plot transformed or raw counts
 #' @param color_pal  a string of a color palette from `khroma` or `RColorBrewer`, or a
 #' vector with colors
 #' @param plot_type one of "violin", "box", or "density" (violin plots, box plots, or
-#' density plots respectively)
+#' density plots respectively). If `plot_meta` and `gene` are used together, then
+#' density plots are disabled
 #' @param ptsize the size of points in the plots
 #' @param ptalpha the transparency of points (violin/box plot) or curves (density plots)
 #'
@@ -28,7 +30,8 @@
 #
 distribution_plots = function(x=NULL, plot_meta=NULL, genes=NULL, samples=NULL, data_type='tr',
                               color_pal='okabeito', plot_type='violin', ptsize=0.5, ptalpha=0.5){
-  #require('ggplot2')
+
+  require('magrittr')
 
   # Define samples to plot if NULL or numeric
   if(is.null(samples)){
@@ -39,7 +42,8 @@ distribution_plots = function(x=NULL, plot_meta=NULL, genes=NULL, samples=NULL,
 
   # Define if plotting meta data or genes
   if(!is.null(plot_meta) & !is.null(genes)){
-    stop("Plots can be generated only for plot_meta OR genes at a time.")
+    plist = cluster_gene_plots(x=x, plot_meta=plot_meta, genes=genes, samples=samples, data_type=data_type,
+                               plot_type=plot_type, color_pal=color_pal, ptsize=ptsize, ptalpha=ptalpha)
   } else if(!is.null(plot_meta)){
     plist = spot_plots(x=x, plot_meta=plot_meta, samples=samples,
                        plot_type=plot_type, color_pal=color_pal, ptsize=ptsize, ptalpha=ptalpha)
@@ -57,7 +61,7 @@ distribution_plots = function(x=NULL, plot_meta=NULL, genes=NULL, samples=NULL,
 # Helpers ----------------------------------------------------------------------
 
 ##
-# violin_plots_spot
+# spot_plots
 #
 
 spot_plots = function(x=NULL, plot_meta=NULL, samples=NULL,
@@ -115,7 +119,7 @@ spot_plots = function(x=NULL, plot_meta=NULL, samples=NULL,
 
 
 ##
-# violin_plots_gene
+# gene_plots
 #
 
 gene_plots = function(x=NULL, genes=NULL, samples=NULL, data_type='tr',
@@ -153,7 +157,7 @@ gene_plots = function(x=NULL, genes=NULL, samples=NULL, data_type='tr',
     gene_cols = color_parse(color_pal, n_cats=length(samples))
 
     # Create plot
-    p_list[[gene]] =
+    #p_list[[gene]] =
     if(plot_type == 'box'){
       p_list[[gene]] = ggplot2::ggplot(df_tmp, ggplot2::aes(x=samplename, y=.data[['geneexpr']], color=samplename)) +
         ggplot2::geom_boxplot(outlier.size=ptsize)
@@ -185,3 +189,94 @@ gene_plots = function(x=NULL, genes=NULL, samples=NULL, data_type='tr',
 }
 
 
+##
+# cluster_gene_plots
+#
+
+cluster_gene_plots = function(x=NULL, plot_meta=NULL, genes=NULL, samples=NULL, data_type='tr',
+                              plot_type='violin', color_pal='roma', ptsize=0.5, ptalpha=0.5){
+
+  if(length(plot_meta) != 1){
+    cat(paste0('Only one metadata variable can be plotted at a time. Plotting ', plot_meta[1], '.'))
+    plot_meta = plot_meta[1]
+  }
+
+  # Check if plot_meta is available and is a discrete variable. Remove sample, otherwise
+  for(i in samples){
+    if(!plot_meta %in% colnames(x@spatial_meta[[i]])){
+      warning(paste0('Variable ', plot_meta, ' is not present in sample ', i, '.'))
+      samples = grep(paste0('^', i, '$'), samples, value=T, invert=T)
+    } else if(!plyr::is.discrete(x@spatial_meta[[i]][[plot_meta]]) | length(unique(x@spatial_meta[[i]][[plot_meta]])) > 30){
+      warning(paste0('Variable ', plot_meta, ' seems continuous or has more than 30 categories and will be skipped for sample ', i, '.'))
+      samples = grep(paste0('^', i, '$'), samples, value=T, invert=T)
+    }
+  }
+  if(length(samples) < 1){
+    stop('All samples skipped.')
+  }
+
+  # Check plot type
+  if(!plot_type %in% c('box', 'violin')){
+    cat("Only 'box' or 'violin' are valid choices. Defaulting to 'box'.")
+    plot_type = 'box'
+  }
+
+  # Extract data slot
+  if(data_type == 'tr'){
+    expr_tmp = x@tr_counts
+    p_title = 'Normalized expression - '
+    ax_title = 'Normalized expression'
+  } else {
+    expr_tmp = x@counts
+    p_title = 'Raw expression - '
+    ax_title = 'Counts'
+  }
+
+  p_list = list()
+  # Loop through list of genes
+  for(gene in genes){
+    # Subset samples if gene not in all samples
+    samples_tmp = samples
+    df_tmp = tibble::tibble()
+    for(i in samples){
+      if(!(gene %in% rownames(expr_tmp[[i]]))){
+        cat(paste0('Sample ', i, ' does not contain ', gene, '.\n'))
+        samples_tmp = grep(paste0('^', i, '$'), samples_tmp, value=T, invert=T)
+      } else{
+        df_tmp = dplyr::bind_rows(df_tmp,
+                                  dplyr::left_join(tibble::tibble(libname=colnames(expr_tmp[[i]]),
+                                                                  geneexpr=expr_tmp[[i]][gene, ],
+                                                                  samplename=i),
+                                                   x@spatial_meta[[i]][, c('libname', plot_meta)], by='libname'))
+      }
+    }
+
+    # Define color palette
+    gene_cols = color_parse(color_pal, n_cats=length(unique(df_tmp[[plot_meta]])))
+    names(gene_cols) = unique(df_tmp[[plot_meta]])
+
+    # Create plot
+    p_list[[gene]] = ggplot2::ggplot(df_tmp, ggplot2::aes(x=samplename, y=.data[['geneexpr']], color=.data[[plot_meta]]))
+    if(plot_type == 'box'){
+      p_list[[gene]] = p_list[[gene]] +
+        ggplot2::geom_boxplot(outlier.size=ptsize)
+    } else if(plot_type == 'violin'){
+      p_list[[gene]] = p_list[[gene]] +
+        ggplot2::geom_violin(ggplot2::aes(fill=.data[[plot_meta]])) +
+        ggforce::geom_sina(size=ptsize) +
+        ggplot2::scale_fill_manual(values=gene_cols)
+    }
+    p_list[[gene]] = p_list[[gene]] +
+      ggplot2::scale_color_manual(values=gene_cols) +
+      ggplot2::ggtitle(paste0(p_title, gene, '\n', plot_meta)) +
+      ggplot2::ylab(ax_title) +
+      ggplot2::xlab(NULL) +
+      ggplot2::theme(axis.text.x=ggplot2::element_text(angle=30, hjust=1),
+                     panel.border=ggplot2::element_rect(fill=NA, color='black')#,
+                     #legend.title=ggplot2::element_blank()
+      )
+  }
+
+  return(p_list)
+}
+

R/filter_data.R

@@ -167,6 +167,8 @@ filter_data = function(x=NULL,
     col_expr_reads = colSums(df_tmp[grepl(spot_pct_expr, rownames(df_tmp)), ])
     col_expr_percent = col_expr_reads/col_total_reads
     rm(col_expr_reads) # Clean env
+    # Force NaNs tp zero, which result from zero as denominator (i.e., zero counts in the cell)
+    col_expr_percent[is.na(col_expr_percent)] = 0
 
     # If no maximum counts set by user, then make it the max counts of each spot/cell
     if(is.null(spot_maxreads)){

R/plot_spatial_meta.R

@@ -81,6 +81,29 @@ plot_spatial_meta = function(x, samples=NULL, ks='dtc', ws=NULL, deepSplit=NULL,
     stop('No metadata column or clustering parameters were specified. Or specified parameters do not exist in metadata.')
   }
 
+  # Set default color if NULL input
+  if(is.null(color_pal)){
+    color_pal = 'light'
+  }
+  # Define color palettes for each meta variable
+  master_col_pal = vector('list', length(plot_meta))
+  names(master_col_pal) = plot_meta
+  for(metacol in plot_meta){
+    cat_vec_tmp = c()
+    for(s in samples){
+      cat_vec_tmp = unique(c(cat_vec_tmp, unique(x@spatial_meta[[s]][[metacol]])))
+    }
+    color_pal_tmp = color_pal
+    if(is.numeric(x@spatial_meta[[s]][[metacol]])){
+      color_pal_tmp = 'sunset'
+    }
+    master_col_pal[[metacol]] = color_parse(color_pal_tmp, length(cat_vec_tmp))
+    names(master_col_pal[[metacol]]) = cat_vec_tmp
+
+    rm(cat_vec_tmp, color_pal_tmp) # Clean env
+  }
+
+  # Make list to store plots a
   plot_list = list()
   for(s in samples){
     # Extract metadata for specific sample
@@ -97,14 +120,6 @@ plot_spatial_meta = function(x, samples=NULL, ks='dtc', ws=NULL, deepSplit=NULL,
     }
 
     for(metacol in plot_meta){
-      # Set default color if NULL input
-      if(is.null(color_pal)){
-        color_pal = 'light'
-        if(is.numeric(x@spatial_meta[[s]][[metacol]])){
-          color_pal = 'sunset'
-        }
-      }
-
       df_tmp2 = df_tmp %>%
         dplyr::select(libname, ypos, xpos, meta:=!!metacol)
 
@@ -115,10 +130,10 @@ plot_spatial_meta = function(x, samples=NULL, ks='dtc', ws=NULL, deepSplit=NULL,
           dplyr::mutate(meta=as.factor(meta))
 
         # Create color palette.
-        meta_cols = color_parse(color_pal, n_cats=length(unique(df_tmp2[['meta']])))
-        names(meta_cols) = unique(df_tmp2[['meta']])
-        if(any(names(meta_cols) == 'No_Data')){
-          meta_cols[names(meta_cols) == 'No_Data'] = 'gray50'
+        # meta_cols = color_parse(color_pal, n_cats=length(unique(df_tmp2[['meta']])))
+        # names(meta_cols) = unique(df_tmp2[['meta']])
+        if(any(df_tmp2[['meta']] == 'No_Data') & any(!grepl('No_Data', names(master_col_pal[[metacol]])))){
+          master_col_pal[[metacol]] = append(master_col_pal[[metacol]], c(No_Data='gray50'))
         }
       }
 
@@ -144,7 +159,7 @@ plot_spatial_meta = function(x, samples=NULL, ks='dtc', ws=NULL, deepSplit=NULL,
         ggplot2::geom_point(ggplot2::aes(x=xpos, y=ypos, color=meta), size=ptsize)
       # Assign color palette to plot for categorical or numerical variable
       if(is.factor(df_tmp2[['meta']])){
-        p = p + ggplot2::scale_color_manual(values=c(meta_cols))
+        p = p + ggplot2::scale_color_manual(values=master_col_pal[[metacol]])
       } else{
         if(!is.null(color_pal)){
           # Get color palette and number of colors needed.
@@ -163,19 +178,18 @@ plot_spatial_meta = function(x, samples=NULL, ks='dtc', ws=NULL, deepSplit=NULL,
 
       if(!is.numeric(df_tmp2[['meta']])){ # Test if it's not numeric and make legend spots/dots larger
         p = p +
-          ggplot2::guides(color=guide_legend(override.aes=list(size=ptsize+1)))
+          ggplot2::guides(color=ggplot2::guide_legend(override.aes=list(size=ptsize+1)))
       }
-      p = p +
-        labs(color=title_leg, title=title_p) + ggplot2::theme_void()
+      p = p + ggplot2::labs(color=title_leg, title=title_p) + ggplot2::theme_void()
 
       if(visium){
         p = p + ggplot2::scale_y_reverse() + ggplot2::coord_fixed(ratio=1)
       } else{
         p = p + ggplot2::coord_fixed(ratio=1)
       }
       #p = p + ggplot2::theme(legend.title=ggplot2::element_blank()) # MAY 09, 2023 PUT META DATA NAME ON LEGEND TITLE, NOT PLOT TITLE
-      p = p + ggplot2::theme(legend.title=element_text(size=txsize),
-                             plot.title=element_text(size=txsize+2))
+      p = p + ggplot2::theme(legend.title=ggplot2::element_text(size=txsize),
+                             plot.title=ggplot2::element_text(size=txsize+2))
 
       plot_list[[paste0(s, '_', metacol)]] = p
     }

R/utils.R

@@ -56,8 +56,7 @@ color_parse = function(color_pal=NULL, n_cats=NULL){
   # Test if input is a Khroma name or RColorBrewer.
   # If so, create palette.
   if(color_pal[1] %in% khroma_cols){
-    p_palette = khroma::colour(color_pal[1], force=T)
-    cat_cols = as.vector(p_palette(n_cats))
+    cat_cols = as.vector(khroma::colour(color_pal[1], force=T)(n_cats))
   }else if(color_pal[1] %in% rownames(RColorBrewer::brewer.pal.info)){
     cat_cols = colorRampPalette(RColorBrewer::brewer.pal(n_cats, color_pal[1]))(n_cats)
   }else{ # Test if user provided a vector of colors.

README.md

@@ -1,5 +1,7 @@
 # spatialGE
 
+<img align="left" src="logo.png" height="100" width="90" />
+
 An R package for the visualization and analysis of spatially-resolved transcriptomics data,
 such as those generated with 10X Visium. The **spatialGE** package features a data object 
 (STlist: Spatial Transctiptomics List) to store data and results from multiple tissue sections,