Biological replicates for my RNAseq experiment

[syntonicC]

Yet another replicates question. I found some other threads here that addressed aspects of my question but I really wanted to make sure it was applicable to my particular scenario:

http://seqanswers.com/forums/showthread.php?t=7905

http://seqanswers.com/forums/showthread.php?t=49586

http://seqanswers.com/forums/showthread.php?t=10813

http://seqanswers.com/forums/showthread.php?t=15574

I am designing a RIPSeq experiment. I would like to see how the population of RNAs immunoprecipitated with my protein of interest changes with the injection of a drug. I also have a protein KO for an extra control in addition to my IgG controls.

My conditions are:

• Tissue from specific brain region, injected with drug, wildtype

• Tissue from specific brain region, injected with vehicle, wildtype

• Tissue from specific brain region, injected with vehicle, knockout

This is a total of three animals, one for each condition. If I perform this in triplicate, I will have a total of 9 animals needed - 3 biological replicates. For each replicate there is the IP sample and the IgG sample for a total of 18 prepped libraries.

I have been told by someone who has published RIPSeq data before that biological duplicates or triplicates are sufficient. However, I have now been told by another researcher that I need a minimum of 6 biological replicates, preferably 8. Doing a quick literature search, I see that two or three replicates have been publishable as of 2015.

Then there are a few papers such as this one:

http://www.ncbi.nlm.nih.gov/pubmed/24319002

...Which suggests more replicates over more reads.

At this point I am not sure what to think. I do not want my paper to be rejected after analysis for having too few replicates. Some of the threads I linked to above are from 2010 or so - have the standards changed since then? I am a little tight on money so my question is: Is it sufficient to use two or three biological replicates for my experiment or do I need more?

If you need any more information about the kind of analysis I will be doing please let me know. Thank you.

[dpryan]

If your budget is tight and 3 replicates is all you can reasonably afford then that will suffice. You will have less power to detect changes and your results will be more influenced by noise, but you'll then need to deal with that. Generally speaking, RIPseq results typically lead to further experiments that will either generally confirm/refute the results, so it's not like you're doing RNAseq and then stopping there.

[syntonicC]

Thank you for your reply.

You will have less power to detect changes and your results will be more influenced by noise, but you'll then need to deal with that.

By this I assume you mean that in the downstream analysis (i.e. DEseq etc.) I will have an opportunity to correct for some of the noise in my data?

Generally speaking, RIPseq results typically lead to further experiments that will either generally confirm/refute the results, so it's not like you're doing RNAseq and then stopping there.

Is qPCR generally the most accepted way confirm/refute results? What are some other approaches commonly acceptable to validate the data? The RIPSeq is actually a relatively small part of my publication and it is being used to supplement behavioral and microscopy data.

[dpryan]

By this I assume you mean that in the downstream analysis (i.e. DEseq etc.) I will have an opportunity to correct for some of the noise in my data?

You won't have any ability to correct for this. The only thing you can do in future experiments based on these results (but not using them!) is hope that you're not wasting too much time/money due to false positives.

Is qPCR generally the most accepted way confirm/refute results? What are some other approaches commonly acceptable to validate the data? The RIPSeq is actually a relatively small part of my publication and it is being used to supplement behavioral and microscopy data.

qPCR is pretty much the standard, since it's cheap to do a bunch of samples. You could also run some Westerns, but qPCR is usually faster.