HW7 - RNA-Seq Analysis

3 minute read

A. Unstranded and strand-specific read counting

• Task 1: Run the RNA-Seq workflow with the toy data from start to finish. Next, rerun the code of the read_counting step here until (including) the line that creates the bfl object with the BamFileList function and then continue with the code shown below. This version performs the read counting the same way as the workflow but lacks the parallelizaton with bplapply.

  In the read quantification step with summarizeOverlaps generate count tables for exons by genes (eByg) of the following three strand modes:

  1. Unstranded
  2. Strand-specific for positive (sense) strand
  3. Strand-specific for negative (antisense) strand

  The solution for generating the unstranded read counts is given below. Note, the upstream code lines of the read counting step in the RNA-Seq workflow only need to be rerun to generate the proper inputs for this taks. Thus, they are not required to be included in the homework results (see HW7.R below).

unstranded <- summarizeOverlaps(eByg, bfl, mode="Union", 
                                                ignore.strand=TRUE, 
                                                # preprocess.reads=invertStrand,
                                                inter.feature=FALSE, 
                                                singleEnd=FALSE)
unstranded <- assays(unstranded)$counts
unstranded[1:4,]

Before attempting to solve this homework task please read the vignette Counting reads with summarizeOverlaps (here) from the GenomicAlignments package that defines the summarizeOverlap function. In addition, the help file for ?summarizeOverlaps provides useful information.

• Task 2: Provide R code that demonstrates that the two strand-specific count tables sum up to very similar values as the unstranded count table.

• Task 3: Explain the utility (biological relevance) of the different strand counting modes used under Task 1. Include your explanation as comment text in your homework script (see HW7.R below).

Note, for Tasks 1-3 only the code and/or text needs to be included in the homework submission (no data/result files). For details see below.

B. Read counting for different feature types

• Task 4: Compute strand-specific count tables for the positive (sense) strand of the following feature types. The help files of ?exonsBy and ?transcripts provide useful information for solving these tasks.

  1. Genes
  2. Exons
  3. Exons by genes
  4. Introns by transcripts
  5. 5’-UTRs by transcripts

Note, for Tasks 4 only include the code and/or text in your homework submission (no data/result files).

C. DEG analysis

• Task 5: Perform the DEG analysis with edgeR as outlined under section 6 of the RNA-Seq workflow here. Use in one case for the DEG analysis the unstranded count table as input (from Task 1.1) and in another the sense strand count table (from Task 1.2). Compare the DEG result of the two methods in two separate 4-way Venn diagrams for the same sample comparisons used in the workflow example here.

  1. 4-way Venn diagram for unstranded count table
  2. 4-way Venn diagram for sense strand count table

Note, for Tasks 5 include both the code and the resulting image files in your homework submission.

Homework submission

Please submit the homework results in one well structured and annotated R script to your private GitHub repository under Homework/HW7/HW7.R, as well as the two Venn diagrams from Task 5. Instead of an R script the homework can be submitted in form of an R Markdown (*Rmd) file. If the latter R Markdown option is chosen, the please include there only the homework solutions, not the entire RNA-Seq workflow itself.

Due date

This homework will be due Tue, May 20 at 6:00 PM.

Homework Solutions

• hw7_solution.R
• Venn diagrams