Batch correction for small RNA-Seq data

Question

[yronimuz]

Hello,

I am new to transcriptomics, I mostly learn by online source and trial and error method. Now, I have 20 samples analyzed for miRNA. I was wondering if I can correct for the batch effects since as you can see below, there is only one sample that contains a unique batch, and for others, they have unequal distributions as well. I already did differential expression analysis using DeSeq2 and edgeR. But these two programs only take raw counts as input. Because by doing so with limma::removeBatchEffect, the data needs to be normalized.

sample  treatment   batch
1   high    4
2   high    6
3   high    6
4   high    6
5   high    6
6   high    7
7   high    8
8   high    9
9   high    9
10  high    9
11  low 6
12  low 7
13  low 8
14  low 9
15  low 9
16  low 9
17  low 9
18  low 9
19  low 9
20  low 9

dds <- DESeqDataSetFromMatrix(countData = cts, colData = meta, design = ~ batch + phenotype)

I was wondering how can I correct for batch and do differential expression analysis with my data? Thank you!

Answer 1

[ATpoint]

What you do with this design is the recommended way of handling batch effects based on the DESeq2 manual. Did you check by PCA of there is evidence for a batch effect?

[yronimuz]

So, if I include the batch on the design, it already corrects for this variable? I did some PCA as shown below:

It seems like there is also no clear clustering between batches so it has no evidence of batch effect. So in the end, I will just include batch on my design and run DeSeq2?

[ATpoint]

From what I understand including batch into the design will correct for the differences between the groups that can be explained by batch (so the unwanted batch variation).

[yronimuz]

I see. Thank you for your replies! Appreciated it.