Question: Multi-factors RNAseq analysis programs
gravatar for qingtian.guan
10 months ago by
qingtian.guan10 wrote:


I am recently analyzing RNAseq data with multiple factors: 2 cell types (A, B), 6 time points (t1, t2, t3, t4, t5, t6), 4 drug concentrations (c1,c2,c3,c4). Normally what I did for one factor RNAseq is to compare TREATMENT to CONTROL and get a list of DE genes. But in this case, I will do so many different combinations:

For example, I can compare the drug treatment effect difference on different cell types: A-t1-c1 compare with B-t1-c1, A-t1-c2 compare with B-t1-c2, A-t1-c3 compare with B-t1-c3, A-t1-c4 compare with B-t1-c4, A-t2-c1 compare with B-t2-c1....etc.

I can also compare the treatment of the drug in different time point, for example. There will different combination and difficult to find the key factor.

I think it won't be too difficult to develop one program that take the data frame structure which contain all the factor information and analysis the DE genes in one go. But I searched and have not found any. Does anyone aware of this kind of program? Thanks.

rna-seq • 240 views
ADD COMMENTlink modified 9 months ago • written 10 months ago by qingtian.guan10

You don't need a special program for this. In R, you can use combn() to get all the possible combinations and then loop through them.

More importantly, as @ATpoint already mentioned, you probably need to better define the question instead of running many combinations.

ADD REPLYlink written 10 months ago by igor11k

Instead of testing every possible contrast I would vote for discussing the experiment in depth involving the PI, the wetlab- and drylab scientists, then define what actually is the major scientific question, and then choose contrasts accordingly. There must have been a basic scientific question when the experiment was performed, so I suggest to work this out and then analyze data to answer it. This will most likely narrow down the choice of the contrasts a lot while eliminating others as they would not contribute.

ADD REPLYlink modified 10 months ago • written 10 months ago by ATpoint41k

Well, this is what DESeq2, edgeR and many more do. You describe the experiment with a formula and it will solve it, you can then look at the different comparisons. As @ATpoint said, you have many factors and you should consider which ones you want to test, which are interacting and how. You will also have to figure out if time and concentration should be treated as discrete or continuous factors.

ADD REPLYlink written 10 months ago by Asaf8.4k
gravatar for qingtian.guan
9 months ago by
qingtian.guan10 wrote:

As Asaf said. I have figured it out now. These multi-factor analysis is included in the DEseq manual. Just need to take care of the experiment design option when using DEseq2. Thank you for your answers. @igor @ATpoint @Asaf

ADD COMMENTlink written 9 months ago by qingtian.guan10
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1224 users visited in the last hour