RNA sequencing data batch effect removal

Question

[Tenghui Chen]

Hi All,

Recently I am trying to analyze some RNA sequencing data and perform the differential expression analysis. Since the sequencing data were generated in different times (experiments), I am worrying about the potential batch effects and would like to find a way to figure out and remove it (remove is more important!)

I tried to process the data in two different ways:

1. Obtain the RSEM values and perform the differential expression analysis, in this way, can anyone kindly provide a good tool to apply on RSEM values to remove the batch effect?
2. Obtain the htseq raw counts and use Deseq2 to perform the differential expression analysis. I saw previously another discussion about about justifying the batch effect by adding the 'batch' in the 'design' [design ~ batch+treatment] parameter in Deseq2. Besides this, can anyone kindly provide another good way to remove the batch effect?

Thanks for the help in advance.

[GouthamAtla]

When you cluster the samples, or do PCA analysis, do you see batch effects as confounding ? I am just curious.

[Tenghui Chen]

Yes, I did PCA, but it did not really show very obvious batch effect, may because I just have very few sample size in each batch (from 2 to 5).

Answer 1

[Ron]

Here is an example using "sva" package:

http://www.bioconductor.org/help/workflows/rnaseqGene/#batch

Remove Batch Effect From RNAseq with SVAseq and Combat

Also,you can use EdgeR.(Example by Devon).Its actually similar to using DESeq.

edgeR: Correct pipeline for DE analysis with multiple conditions and batches

[Tenghui Chen]

Thank you. I do not actually quite understand how to achieve a best practice with SVAseq. Do you know how to decide the n.sv value in the function?