Nice Interactive Analysis in R package

Here I collect many R functions that I’ve created over time. The primary purpose of this package is to neatly organise my R code in one place and have it ready to use whenever. Plus this forces me to document my functions and it's easier to share the analysis code. The code could be subject to frequent changes and is "always" in development. Feedback is welcome! If you find an error please open a GitHub issue.

Requirements

• R >=4.1.0
• Mac or Linux operating system (not tested on Windows)

Installation

There are 2 options to use this package.

From this GitHub Repository

If interested in giving it a try:

devtools::install_github("Ni-Ar/niar")

Load the local repository (only for CRG users)

If you have a CRG user account, instead of installing, you can load the R package at the beginning of your session. Log in with your credentials on the CRG RStudio Server IDE (R version 4.3.3) and run:

devtools::load_all(path = '/users/mirimia/narecco/software/R/niar')

This command will just load the package (from my local repository on the CRG cluster). For this step to work properly you must make sure that all the required dependencies have been already installed as just loading the package won’t install the dependencies for you. You can install the required packages with: (this step takes a while)

install.packages(c('devtools', 'matrixStats', 'BiocManager', 'XICOR', 
                   'ggplot2', 'ggrepel', 'scales', 'patchwork',
                   'MetBrewer', 'ggalluvial', 'ggfittext', 'ggseqlogo', 'seqinr',
                   'dplyr', 'tidyr', 'tibble', 'forcats', 'stringr'))
install.packages('Cairo') 
# if you get an error installing 'Cairo' you might need to first install the cairographics C library your operating system from https://www.cairographics.org/download/

and the following Bioconductor packages:

BiocManager::install("Biostrings")
BiocManager::install("biomaRt")
BiocManager::install("DESeq2")
BiocManager::install("csaw")
BiocManager::install("msa")

If you get an error with this method it’s probably cause by the fact that the dependencies are already in the .libPaths(). Remember that once you log out of your R session the niar functions won’t be availbale anymore and you’ll need to load them again next time.

Troubleshooting plots in RStudio server

In order to visualise the plots you might need to select the right graphics device, especially if you get an error that says something like:

Error in diff.default(from) : 
  Shadow graphics device error: r error 4 (R code execution error)
In grDevices:::png("/tmp/Rtmp....",  :
  unable to open connection to X11 display ''

To solve this go to: Tools Menu (on top of the windos) > Global Options > General section > Graphics tab > and in the Graphic Device Backend drop down menu select Cairo and then click “Apply”. Now the plots should be correctly displayed.

Overview

Currently this package contains:

• one function to perform Principal Component Analysis (PCA) in 2D with lots of option to enrich visualisation and exploration. See the vignette below for more details.
• several functions to fetch and parse data analysed with vast-tools for alternatively spliced events and gene expression. There are also plotting functions to quickly glimpse into the data (e.g. plot_corr_gene_expr_psi()).
• Some publically available datasets have been packaged in ad-hoc funtions to quick plot and explore the data:
  • Mouse Development ENCODE data plot_mouse_tissue_devel() which uses data I preprocessed fetched with get_mouse_tissue_devel_tbl(). Currently works only on the CRG RStudio Server IDE.
• Some Biomart handy functions for quick gene IDs conversions (e.g. ensembl_id_2_gene_name()).
• Some DESeq2 wrappers.
• Multiple sequence analysis from fasta format to generate PWMs, visualise logos or Jensen-Shannon divergence.

More examples grouped by topic are listed below:

PCA

The easiest way to make a PCA assuming mat is your numerical matrix is:

showme_PCA2D(mat)

To know more you can type:

?showme_PCA2D()

The underlying function is prcomp and you can pass extra arguments with ... for example:

showme_PCA2D(mat, scale. = T, center = F)

Extra info can be passed from a metadata dataframe with mt =. To specify which column of the dataframe contains the colnames of the matrix mat use mcol. In the following example the mt contains a column called sample_name:

showme_PCA2D(mat = mat, mt = mt, mcol = "sample_name", show_variance = T, show_stats = T)

To show the PCA loadings:

showme_PCA2D(mat = mat, n_loadings = 12)

More details can be found in the vignette below.

vast-tools

Since I use vast-tools quite often I made functions to easily import the output tables into R. Namely, grep_psi() or grep_gene_expression() import the PSI of an AS events or gene expression levels respectively and parse the data into a long-format dataframe with the accompanying tidy functions tidy_vst_psi() or tidy_vst_expr(). These functions work great with the magrittr pipe (%>%) or the base R pipe operator (|>) as in:

grep_psi(inclusion_tbl = file.path(dir_location, "INCLUSION_LEVELS_FULL-hg38-n-v251.tab"), 
         vst_id = c("HsaEX0000001", "HsaEX0000002")) |>
    tidy_vst_psi() 

These functions are basically “hacks” that call the system grep command, write to a temporary file that is then read into R and removed from the system. Maybe a better way would probably be to implement the functions in Rcpp.

Vignettes

Link for PCA.

Link for mouse ENCODE AS data AS exploraion.

To do

• Make vignette for biomaRt functions
• Make vignette for vast-tools utility and plotting functions, especially correlations.
• Maybe add the mouse ENCODE data (fetched with get_mouse_tissue_devel_tbl) to the package?
• Make vignette for logo analysis