Normalizing For Batch Effect In RNA-Seq Data

Question

[Tom A]

Hello,

Does anyone have a suggestion for dealing with "batch effect" between two batches of small RNA-Seq data? I've heard "ComBat" was good. Anyone have any others in mind?

Thanks in advance.

[Ahill]

This paper is worth a look. They tried several methods on array data (including "small" sample sizes), and concluded ComBat performed best: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0017238

Answer 1

[Ann]

The EdgeR package in Bioconductor can help with this. See the Arabidopsis data set in the User's Guide where samples were collected at different times.

[Tom A]

Thank you. I will give it a look over.

Answer 2

[tonja.r]

Try RUVseq, they have three different approaches how to deal with batch effects based on empirical genes, in-spike genes and samples contrast. Also, take a look on PEER.

Answer 3

[Ron]

I used combat and it does it pretty well.