Question: Compare groups of different RNAseq sets
0
gravatar for b.nota
18 months ago by
b.nota340
Netherlands
b.nota340 wrote:

I would like to compare two groups (cell populations), from two different RNAseq experiments (sets). However, there are no overlapping groups in both experiments available, so an analysis in limma with batch effect (block design) is not possible. Let's say I have populations a, b, and c in experiment 1, and d, and e in experiment 2, and I would like to compare a with d.

Is there another way to compare the two populations in a sound manner? Or is it only possible to make a heatmap? And compare average cpm values?

limma batch effect • 299 views
ADD COMMENTlink modified 18 months ago by Gordon Smyth38k • written 18 months ago by b.nota340
Answer: Compare groups of different RNAseq sets
1
gravatar for Aaron Lun
18 months ago by
Aaron Lun25k
Cambridge, United Kingdom
Aaron Lun25k wrote:

There is no way to compare groups a and d, either in a DE analysis or via a heatmap or via comparisons of average expression. If there are any batch effects, comparisons across batches of any type will be compromised - this includes informal visual comparisons. The solution? Design a better experiment.

ADD COMMENTlink modified 18 months ago • written 18 months ago by Aaron Lun25k

Thanks, I understand that. Just to be clear the experiments were both independent, with good design. But an additional question arose how a group in experiment 1 is compared to another group in a previous experiment. I mean what is the use of GEO data in this case?

ADD REPLYlink written 18 months ago by b.nota340

You can't use GEO data like that. You can do meta-analyses, but that involves comparisons of log-fold changes within batches.

ADD REPLYlink written 18 months ago by Aaron Lun25k
Answer: Compare groups of different RNAseq sets
1
gravatar for Gordon Smyth
18 months ago by
Gordon Smyth38k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth38k wrote:

The way to robustly compare different experiments is by gene set tests. For example, compare d vs e using experiment 2. That gives a list of DE genes with associated log-fold-changes, which we might call the d vs e expression signature. Then use a roast gene set test to examine whether the d vs e signature is up-regulated when you compare a vs b and a vs c using experiment 1.

This method works because it depends on comparisons that are made within each individual experiment --- at not stage are conditions compared directly across batches.

I use this technique in many of my own published papers, if you want to have a look for examples.

ADD COMMENTlink modified 18 months ago • written 18 months ago by Gordon Smyth38k

This sounds like a good solution, many thanks. I am going to give that a try!

ADD REPLYlink written 18 months ago by b.nota340
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 198 users visited in the last hour