In systemPipeR/systemPipeTools: Tools for data visualization

```{css, echo=FALSE, eval=TRUE} pre code { white-space: pre !important; overflow-x: scroll !important; word-break: keep-all !important; word-wrap: initial !important; }

```r
BiocStyle::markdown()
options(width = 80, max.print = 1000)
knitr::opts_chunk$set(
  eval = as.logical(Sys.getenv("KNITR_EVAL", "TRUE")),
  cache = as.logical(Sys.getenv("KNITR_CACHE", "TRUE")),
  tidy.opts = list(width.cutoff = 80), tidy = TRUE
)

suppressPackageStartupMessages({
    library(systemPipeTools)
    library(systemPipeR)
})

Data Visualization with `systemPipeR`

systemPipeTools package extends the widely used systemPipeR (SPR) [@H_Backman2016-bt] workflow environment with enhanced toolkit for data visualization, including utilities to automate the analysis of differentially expressed genes (DEGs). systemPipeTools provides functions for data transformation and data exploration via scatterplots, hierarchical clustering heatMaps, principal component analysis, multidimensional scaling, generalized principal components, t-Distributed Stochastic Neighbor embedding (t-SNE), and MA and volcano plots. All these utilities can be integrated with the modular design of the systemPipeR environment that allows users to easily substitute any of these features and/or custom with alternatives.

Installation

systemPipeTools environment can be installed from the R console using the BiocManager::install command. The associated package systemPipeR can be installed the same way. The latter provides core functionalities for defining workflows, interacting with command-line software, and executing both R and/or command-line software, as well as generating publication-quality analysis reports.

if (!requireNamespace("BiocManager", quietly = TRUE)) 
  install.packages("BiocManager")
BiocManager::install("systemPipeTools")
BiocManager::install("systemPipeR")

Loading package and documentation

library("systemPipeTools")  # Loads the package
library(help = "systemPipeTools")  # Lists package info
vignette("systemPipeTools")  # Opens vignette

Metadata and Reads Counting Information

The first step is importing the targets file and raw reads counting table. - The targets file defines all FASTQ files and sample comparisons of the analysis workflow. - The raw reads counting table represents all the reads that map to gene (row) for each sample (columns).

## Targets file
targetspath <- system.file("extdata", "targets.txt", package = "systemPipeR")
targets <- read.delim(targetspath, comment = "#")
cmp <- systemPipeR::readComp(file = targetspath, format = "matrix", 
                             delim = "-")
## Count table file
countMatrixPath <- system.file("extdata", "countDFeByg.xls",
                               package = "systemPipeR")
countMatrix <- read.delim(countMatrixPath, row.names = 1)
showDT(countMatrix)

Data Transformation

For gene differential expression, raw counts are required, however for data visualization or clustering, it can be useful to work with transformed count data. exploreDDS function is convenience wrapper to transform raw read counts using the DESeq2 package transformations methods. The input file has to contain all the genes, not just differentially expressed ones. Supported methods include variance stabilizing transformation (vst) (@Anders2010-tp), and regularized-logarithm transformation or rlog (@Love2014-sh).

exploredds <- exploreDDS(countMatrix, targets, cmp = cmp[[1]], preFilter = NULL, 
                         transformationMethod = "rlog")
exploredds

Users are strongly encouraged to consult the DESeq2 vignette for more detailed information on this topic and how to properly run DESeq2 on datasets with more complex experimental designs.

Scatterplot

To decide which transformation to choose, we can visualize the transformation effect comparing two samples or a grid of all samples, as follows:

exploreDDSplot(countMatrix, targets, cmp = cmp[[1]], preFilter = NULL,
               samples = c("M12A", "M12A", "A12A", "A12A"), 
               scattermatrix = TRUE)

The scatterplots are created using the log2 transform normalized reads count, variance stabilizing transformation (VST) (@Anders2010-tp), and regularized-logarithm transformation or rlog (@Love2014-sh).

Hierarchical Clustering Dendrogram

The following computes the sample-wise correlation coefficients using the stats::cor() function from the transformed expression values. After transformation to a distance matrix, hierarchical clustering is performed with the stats::hclust function and the result is plotted as a dendrogram, as follows:

hclustplot(exploredds, method = "spearman")