Skip to contents

Setting Up Your R Environment

brentlabRnaSeqTools.Rmd

Requirements

An up to date R (so whatever the current version is – check CRAN) and up to date Rstudio

Install the following 


cran_pkgs = c('tidyverse', 'devtools')

# note: increase/decrease the number of CPUs. Any machine should be able to 
# handle at least 5. You can probably bump this up to 10 without checking
# anything
install.packages(cran_pkgs, Ncpus = 5)

It is possible that there wil be install errors that come up during the installation of these two pages. If you are on the computational/analyst side of things, then this is a good opportunity to read the error messages and figure out what to install – both packages may have some c or maybe fortran dependencies which you will need to install onto your computer. Googling the error helps, and frequently R will say, “you need a gcc compiler” or something like that.

If you are not one of the computational/analyst people, then ask for help if there are errors in the installation of those two packages.

Install the brentlabRnaSeqTools

If there is a ‘passing’ badge in the github README, then this has been successfully built on up to date linux, mac and windows OS. It should also work for you. But, if it doesn’t, copy the error and make an issue report. Please also include the output of Sys.info() in your issue report.

# as above, you can increase Ncpus
remotes::install.packages("BrentLab/brentlabRnaSeqTools", Ncpus = 5)

Set up some environmental variables

Environmental variables are variables that are read by R and loaded into your session when you launch R. You can set either ‘user’ level variables, which are loaded into any R session launched under your current user, and/or you can set project level environmental variables, which are only set if you launch into a project (see Using Rstudio Projects below). These are convenient to use, and are particularly good at avoiding any embarrassing leaks of login credentials onto github.

User level environmental variables

  1. open Rstudio
usethis::edit_r_environ('user')
  1. in the file that is opened, enter some variables that you’d like to have
db_username = "some_username"
db_password = "some_password"

  1. restart your R session

  2. see that you can access these variables like so:

Sys.getenv("db_username")
# output will be:
#> "some_password"

If you put the correct username/password into your environment file, then you can access the database like so:

library(brentlabRnaSeqTools)

meta = getMetadata(
  database_info$kn99$db_host,
  database_info$kn99$db_name,
  Sys.getenv("db_username"),
  Sys.getenv("db_password"))

Since the user level environment file is stored in ~/.Renviron, there is no danger of pushing these credentials up to git if you are in a directory which you are tracking with git/github.

Using an Rstudio project

set your working directory

If you do not already have a directory in which you store the projects you work on, create one. I suggest a direcory called projects in your $HOME

project_dir = "~/projects"
if(!dir.exists(project_dir)){
  dir.create(project_dir)
}
setwd(project_dir)

Create a new Rstudio project

A project is a subset of the infrastructure in an R software package. You can use a virtual environment, .Renviron files for environmental variables, etc. It is a tool for reproducibility

# if this is for the nineyMinuteInduction data, for example, you might call 
# this project "ninetyMinInduction".
project_name = "my_new_project"

usethis::create_project(project_name)

Using the project directory

A new Rstudio session will launch in your new project directory. Now, whenever you launch this project, all of your environment variables in .Renviron will be loaded. If you are using a virtual environment (a good idea for reproducibility. Use renv), then your virtual environment will be automatically launched, also.

The /R directory is for R scripts. You could make a notebooks directory, or just put the notebook in the project parent directory, for example.

See here for a project directory example

Using project level environmental variables in an active R project

Just as with the user level environmental variables, you can set environmental variables for a particular project. These are read in addition to the user level variables, and will overwrite them if there are two that are named the same thing. The first thing to do is to make sure that the project level .Renviron is in the .gitignore so that you don’t accidently push up login credentials

# you can use this, or just click on the .gitignore file in the project
usethis::edit_git_ignore('project')

In the .gitignore file, add .Renviron. Next, open a project level .Renviron

usethis::edit_r_environ("project")

and edit as before. Remember to re-launch your R session after editing the .Renviron to have access to the environmental variables.

Using data from the lts archive 

library(brentlabRnaSeqTools)
library(tidyverse)

# mount the cluster, or download the 20220208 kn99 db archive
# paths will need to be updated for your machine. These are examples
# of what it would look like if you mount.
archive_prefix = '/mnt/lts/sequence_data/rnaseq_data/kn99_database_archive'
coldata_df = read_csv(file.path(archive_prefix, "20220208/combined_df_20220208.csv"))
count_df = read_csv(file.path(archive_prefix, "20220208/counts.csv"))

gene_ids = getGeneNames(database_info$kn99$db_host,
                        database_info$kn99$db_name,
                        Sys.getenv('db_username'),
                        Sys.getenv('db_password'))

gene_ids = gene_ids$gene_id[1:6967]

coldata_df_fltr = coldata_df %>%
  filter(purpose == "fullRNASeq",
         !is.na(fastqFileName))

count_df_fltr = count_df[, colnames(count_df) %in% 
                           coldata_df_fltr$fastqFileName]

coldata_df_fltr = filter(coldata_df_fltr, 
                         fastqFileName %in% 
                           colnames(count_df_fltr))

# make sure the order of the count cols matches order of the metadata rows
count_df_fltr = count_df_fltr[order(match(colnames(count_df_fltr),
                                            coldata_df_fltr$fastqFileName))]

# if this checks false, stop and figure out why. useful functions would be 
# setdiff(). Read the ?setdiff docs -- it is asymetric.
stopifnot(identical(colnames(count_df_fltr), 
                    coldata_df_fltr$fastqFileName))

dds = DESeq2::DESeqDataSetFromMatrix(countData = count_df_fltr,
                                     colData = coldata_df_fltr, 
                                     design = ~1)
rownames(dds) = gene_ids

blrs = brentlabRnaSeqSet(dds = dds)

# everything should work as usual from here
createEnvPert_epWT = function(blrs){

  # filter
  ep_meta_fastqFileNumbers = extractColData(blrs) %>%
    filter(strain == 'TDY451',
           treatment %in% c("cAMP","noTreatment"),
           treatmentConc %in% c("20", 'noTreatmentConc'),
           experimentDesign == 'Environmental_Perturbation',
           purpose == "fullRNASeq",
           libraryProtocol == 'E7420L',
           !is.na(fastqFileName)) %>%
    pull(fastqFileNumber)

  ep_set = blrs[,colData(blrs)$fastqFileNumber %in%
                  ep_meta_fastqFileNumbers]

  experimental_conditions = alist(medium,
                                  atmosphere,
                                  temperature,
                                  treatment,
                                  treatmentConc,
                                  pH,
                                  timePoint)

  concat_treatment = extractColData(ep_set) %>%
    mutate(concat_treatment = as.factor(paste(!!!experimental_conditions,
                                              sep="_"))) %>%
    pull(concat_treatment)

  colData(ep_set)$concat_treatment = concat_treatment

  ep_set

}