Skip to content

Analysis of differential gene expression in PRJNA630433 using edgeR

edgeR Analysis

To begin, navigate to the history PRJNA630433 FeatureCounts Counting on HISAT2 bam alignments and copy the three dataset collections of counts generated by FeatureCounts: Dc FeatureCounts counts, Mo FeatureCounts counts and Oc FeatureCounts counts into a new history that you will name PRJNA630433 edgeR analysis

Then, search for edgeR in the tool search bar

edgeR settings

  • Count Files or Matrix?

    → Separate Count Files

  • 1: Factor/Name

    → Tissue

  • 1: Factor/1: Group

    Note that there will be three Groups (ie factor levels) in this analysis: Dc, Mo and Oc.

    → Oc

  • Counts file(s)

    → select the data collection icon, then 15: Oc FeatureCounts counts

  • 2: Factor/2: Group

    → Mo

  • Counts file(s)

    → select the data collection icon, then 10: Mo FeatureCounts counts

  • 3: Factor level (you must click on ➕ Insert Group)

    → Dc

  • Counts file(s)

    → select the data collection icon, then 5: Mo FeatureCounts counts

  • Use Gene Annotations?

    → No

  • Formula for linear model

→ Leave empty - Input contrasts manually or through a file

1
--> `manually`

  • 1: Constrast

    → Mo-Dc

  • 2: Constrast (click ➕ Insert Contrast)

    → Oc-Dc

  • 3: Constrast (click ➕ Insert Contrast)

    → Oc-Mo

  • Filter Low Counts

    → No, leave folded

  • Output options

    → Leave folded

  • Advanced options

    → Put P-Value Adjusted Threshold to 0.1 (to be consistent with DESeq settings)

    → Leave other advanced options unchanged

  • Run Tool

Note on the order of Factors levels (Groups) in the edgeR html form

In contrast to DESeq2, the order of the Factors levels (Groups) does not matter with the edgeR approach.

This is because here you specify manually the comparison formulas. Yet, in these formula, the order of the levels matters !

Thus when we specify Mo-Dc this implies specifically that we consider the Dc as the reference level: we "subtract" the test level Mo from the reference level Dc

Inspect edgeR plots

There is a lot of information here which we will discuss online or in live. You should also compare these plots side by side with the plots generated by DESeq2.

Here, no need for adding header to edgeR tabular outputs !