A question about using HISAT2 without biological replicates

Hi all,

I have some RNA-Seq data to analyze, but unfortunately, we only have one biological replicate. I therefore analyzed with the Tuxedo pipeline and got a list of DEGs. The samples here are untreated and treated mice, and the number of DEGs is only coming to about 150, and none of the targets of our interest are showing up. I am guessing that the treatment has not worked.

I was suggested to try this with the HISAT2 pipeline, but when I was reading through it, I learned that we need biological replicates for it. My question is since the data is paired-end, will it be a good option to provide the mate pairs as technical replicates for the analysis, and if not, why?

Thanks in advance

rna-seq hisat2

• 1,320 views • link

written 2.5 years ago by vinayjrao

1 answer

Yes, you need biological replicates for any experiment in any setup. That's a basic of experimental design in any discipline.

You cannot use the mates as they're from the exact same experiment. This would be cheating / making up replication where is none.

You can read the edgeR documentation for an unreliable workaround to 'simulate' replication to get a rough idea of DEGs.

But actually you need to repeat the experiment with replication to have robts results. Anyway, if the fold changes of all interesting genes indicate that the treatment did not work you might include that into decision whether to repeat.

My advise is to look into analysis and talk to an analyst before you do and plan an experiment.

• 0 views • link

written 2.5 years ago by ATpoint

Yes, and thank you so much for your response. I had already informed my colleague about the importance of biological replicates for this case, but only since cuffdiff calculates p-values even in the case of no biological replicates, that method was followed for downstream analysis.

However, upon identifying only a few dysregulated genes, I was asked to try the analysis with hisat2, and when I again explained the issue to them, I was asked to do it while using the forward and reverse reads as replicates. I hope the view of another person will convince them otherwise.

Thanks for keeping it simple :)

• 0 views • link

written 2.4 years ago by vinayjrao