Filtering gene counts prior to edgeR DEA

[ChristopherAdelmann]

Operating System

Ubuntu 22.04

Other Linux

No response

Workflow Version

v1.7.2

Workflow Execution

Command line (Local)

Other workflow execution

No response

EPI2ME Version

No response

CLI command run

No response

Workflow Execution - CLI Execution Profile

None

What happened?

Thank you for providing this workflow.
This is a follow-up question/issue to #57

While reviewing some DE-Analysis results, I stumbled upon multiple questionable genes that are significantly enriched/depleted. For example, these genes had high counts in only one of the triplicates.

In the previous issue, you (@sarahjeeeze) mentioned that only the DTU analysis utilizes the filtered genes list, and that one should use a pre-filtered list for DEA. While parsing the workflow code, I couldn't figure out how to do this, since the pipeline uses the salmon quantification output directly. Is there any particular reason why not to use edgeR's built-in filtering method as described in the vignette.
Filtering of gene counts prior to edgeR analysis as described in the vignette 2.7.

Excerpt of vignette:
In addition, the pronounced discreteness of these counts interferes with some of the statistical approximations that are used later in the pipeline. These genes should be filtered out prior to further analysis

Thanks for clarifying.

Best regards,
Christopher.

Relevant log output

Not relevant.

Application activity log entry

Were you able to successfully run the latest version of the workflow with the demo data?

yes

Other demo data information

[sarahjeeeze]

Hi, thanks for reporting this. What i should of said is the input parameters are used for DTU specifically for the transcript level filtering used by drimseq. But you are right we should additionally be filtering before edgeR, we will incorporate this feedback in a future release. In the meantime you can edit the de_analysis.R script to add a filtering step filterByExpr().