DESeq2, how to deal with batch effect: t-test or design matrix?
Entering edit mode
Mozart ▴ 20
Last seen 3.0 years ago

Hello, generally speaking, when we are dealing with human tissues dataset (different hospitals, different days, etc.) we have to take into account the so-called batch effect. I have been practising with some useful tool such as SVA package but someone else like to cope with this issue in a slightly different way: either by doing a t-test on the whole dataset and designing a matrix.

I have always wondered why doing a t-test might control the batch effect, someone more expert than myself said that this is not so robust. It would be better to directly design a matrix...but it seems very hard for my beginner experience. But I guess it is something I have to learn because, for the kind of experiment I am doing (either on humans and mice) it is impossible to perform the experiment on the same day.

If someone could help me to understand:

  1. if the design matrix is the best way to control batch effect,
  2. how much we lose in term of reliability by doing a t-test (rather than a normal DESeq2 analysis)
  3. if tools like SVA package can control the batch effect better than a design matrix written from scratch
  4. if someone could also share the code from the DESeq2 package that allows to perform a t-test (either paired and unpaired) would really help me a lot since I was not able to find it.

Regards, Maria

deseq2 batch sva t-test design matrix • 646 views
Entering edit mode
Last seen 13 hours ago
United States

DESeq2 is a package for analysis of count matrices for high throughput sequencing experiments. The t test isn’t really an alternative or an option within DESeq2.

Maybe read over the DESeq2 publication and then the beginners workflow and see if you have further specific questions.


Login before adding your answer.

Traffic: 658 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6