Hourglass
Tool for visualization of numeric data with annotations
The goal of Hourglass is to quickly acquire color vectors (named palettes and colors for heatmaps/correlations/any scaled values) via the RColorBrewer package and colorRampPalette function. See Documentation {link} for more information.
Installation
You can install the latest version from Github:
devtools::install_github("kazeera/Hourglass", upgrade_dependencies = FALSE)
How to Use
Load package.
library(Hourglass)
There are 4 components: 1) comparisons and user options 2) color palette for subgroups 3) feature sets and respective parameters 4) optional - dataset object
Option 1: Run from Hourglass parameters Excel file.
Download {link} for template.
run_from_excel("path_to_file/file.xlsx")
Option 2: Run from individual components.
run_Hourglass(comparisons, var_colors, feat_sets, datasets)
Example of features and functionality.
Data preparation
Define dataset object - list with 3 tables: 1) "vals" (sample by feature/parameter matrix) 2) "rowAnn" (row annotations = sample/patient information) 3) "colAnn" (column annotations = feature/parameter information)
Load in example dataset object.
# Retrieve built in dataset ds <- example_IHC_samples_dataset # Name your dataset ds$name <- "BySample"
Note: use make_dataset_ob() to create your own.
Remove outliers.
ds$vals <- remove_outliers_df(ds$vals)
Optional: Create new imputed dataset to run in parallel.
ds.imp <- impute_ds(ds)
Add IL6 levels to rowAnn. Leverage quantified resultsto create stain-based annotation column.
# Get levels from "IL6_Num.Positive" result <- add_to_rowAnn(ds, col_name = "IL6_Num.Positive") # Add new rowAnn with new column to ds ds$rowAnn <- result$rowAnn result$rowAnn1 # [1] "IL6_Num.Positive" unique(ds$rowAnn[,"IL6_Num.Positive"]) # [1] "intermed" NA "high" "low"
Create color palette list for variables/subgroups.
var_colors <-list('Sample_Cancer_Subtype-A'="#9a8aabff",'Sample_Cancer_Subtype-B'="#9446e8ff", 'Sample_Cancer_Subtype-C'="#3b3374ff", 'Sex-F'="#bff2A7ff",'Sex-M'="#f64d47ff", 'Smoker-Yes'="#d6c938ff",'Smoker-No'="#3e8a5bff", 'Custom-low'="#18528cff",'Custom-intermediate'="#ebb7b7ff",'Custom-high'="#911919ff") var_colors <- clean_colors(color_palette) # Show color palette using scales package scales::show_col(unlist(var_colors))
Note: use show_col() from scales package to show palette (vector of color HEX codes/names)
Use individual plotting functions.
Boxplot - Comparing cancer subtypes.
Make a single boxplot. Notice how the differences are siginificant.
library(gtools) library(ggplot2) # Prepare 2 column data frame df3 <- data.frame(box = ds$rowAnn$Sample_Cancer_Subtype, value = ds$vals$CD20_Pos.Pixel.Percent) # Boxplot 2 plot_indiv_boxplot(df3, labels = "CD20 in Cancer Subtypes", ylab = "CD20_Pos.Pixel.Percent", x = "", lvl.colors = c(A="#9a8aabff",B="#9446e8ff", C="#3b3374ff"), save.to.file = F)
Make another boxplot for IL6 which shows no significance.
# Prepare 2 column data frame df3 <- data.frame(box = ds$rowAnn$Sample_Cancer_Subtype, value = ds$vals$IL6_Num.Positive) # Boxplot 1 plot_indiv_boxplot(df3, labels = "IL6 across Cancer Subtypes", ylab = "IL6_Num.Positive", lvl.colors = c(A="#9a8aabff", B="#9446e8ff", C="#3b3374ff"), save.to.file = F)
Try again but only look at female patients. The trend is now visible and significant now (p < 0.05).
library(gtools) library(ggplot2) # Subset dataset object ds_F <- subset_dataset(ds, rows_to_keep = ds$rowAnn$Sex == "F") # Prepare 2 column data frame df3 <- data.frame(box = ds_F$rowAnn$Sample_Cancer_Subtype, value = ds_F$vals$IL6_Num.Positive) # Boxplot 1 plot_indiv_boxplot(df3, labels = "IL6 in Cancer Subtypes in female", ylab = "IL6_Num.Positive", lvl.colors = c(A="#9a8aabff",B="#9446e8ff", C="#3b3374ff"), save.to.file = F)
Note: You may add a second subgroup annotation.
# Make a third column df3$dots <- ds_F$rowAnn$Smoker # Boxplot 3 - add more colors to color palette "lvl.colors" plot_indiv_boxplot(df3, labels = "Comparing IL6 in Sample Cancer Subtypes", lvl.colors = c(A="#9a8aabff",B="#9446e8ff", C="#3b3374ff", 'Yes'="#d6c938ff",'No'="#3e8a5bff"), save.to.file = F)
Feature sets - correlation plots and heatmap
Survival plot
Make survival plot to see differences between sexes.
library(survival) # Make table with 3 columns df1 <- data.frame(time = ds$rowAnn$OS_time, status = ds$rowAnn$Status, col = ds$rowAnn$Sex) # Plot survival curve plot_surv_curve(df1, label = "Sex", out_dir = ".")
