How To Combine Technical And Biological Replicates In Rna-Seq Studies?

Question

[camelbbs]

Hi,

I want to ask in analysis of rnaseq data to detect differential expressed genes, etc., can technical replicates and biological replicates put together to compare?

Thanks,

Ch

Answer 1

[Steve Lianoglou]

The standard recommendation is to combine your technical replicates into one "lane" by simply adding their reads together. Use your biological replicates to estimate variance/dispersion.

[camelbbs]

Since I don't have biological replicates for one sample, so I want to use technical rep for analysis, I don't know if that is also working

[Steve Lianoglou]

If you have, say two conditions, with 2 biological replicates in condA but no biological replicates in condB, there are ways to accommodate for this. Read through the edgeR and DESeq vignettes to get an idea. The answer is not to use your second technical replicate as a biological one.

[camelbbs]

But I am not sure if the pvalue from partial replicate samples are believable

[Steve Lianoglou]

Then sequence another biological replicate?

I'm not sure where you want this conversation to go ... I'm just pointing you to where I think you'll find a reasonable thing to do with your data explained within a tutorial.

[camelbbs]

Thanks a lot! I think you gave very reasonable suggestions.

[ccshao]

A little bit confused here. "simply adding their reads together" means add the read counts after tophat and htseq or merge the raw data (fastq) into a single file, then go through tophat and htseq ?

[camelbbs]

means add the read counts

[Steve Lianoglou]

If you already aligned both datasets separately, merging the two bams together then doing feature counting via ht-seq on the merged bam would get you the exact thing I'm suggesting to do if you have technical replicates. There are other ways to get equivalent counts, and if those are easier for you to do, then do that but hopefully you get the basic idea.