I have some RNA-seq data, in which some samples were paired-end reads, and some were single-end reads. I used Combat-seq from sva package to remove the batch effect with input is raw counts. Then I used the adjusted counts (after Combat-seq) as input for DESeq2 and made PCA plot (rld).
I also compiled some other RNA-seq data (paired-end) from 3 papers (I make a raw-count file again from their raw RNA-seq). Now, I set 5 batches: 1: my paired-end data, 2 my single-end data, 3-5: data from 3 different papers. Is that fine?
I am trying to use other methods of batch effect removal and compare those. I tried using the limma::removeBatchEffect. I did make dds file from DESeq2 (input is raw counts) => vsd (<-vst function) => limma::removeBatchEffect => PCA plot.
I had 2 PCA plot show totally differently by 2 different methods of batch effect removal.
I am so new to bioinformatics. Please give me some advice or share with me some relevant materials.
Thank you in advance!
1 answer
Sequencing techniques, platform and different sources of data should be considered as independent confounders. Hence for your expression, it should be something like: count~end+platform+cohort.
Log in to answer this question.
-
Batch and Sample correction for downstream analysis using DESeq2
written by NorbertK •
-
DESeq2 Batch correction enquiry?
written by prithvi.mastermind •
-
limma::removeBatchEffect and using DESeq2 for RNA-seq differential expression
written by 2405592M
-
Batch correction in DESeq2
written by Arindam Ghosh
-
Batch effect and DE using TCGA and my data
written by thaisenayane •
-
Correcting for batch effect in RNA-seq data
written by Rimma •
-
Weird results after batch effect removal using ComBat.
written by goldie.ed •
-
Individual genes plots after batch correction with Deseq2
written by Eugene A
-
Pre-processing of RNA-Seq data for WGCNA
written by gokce.ouz
-
Evaluating removing batch effect tools with PCA or MDS?
written by Lluís R.
