Entering edit mode

pbachali
▴
50

@pbachali-9651
Last seen 4.5 years ago

Hi all,

I have a good amount of experience in doing differential gene expression analysis in two categories like healthy samples and disease patients. Now I have three categories of patients at baseline, then at week 16 and week 52. I am thinking to try one way anova on this data. Any information on links or any good sources of DE analysis with anova awould be really appreciated.

Thanks,

Prat

You say you have a "good amount of experience" with DE analysis. What software have you used to do these analyses? I am wondering why anova causes such a special difficulty.

I have been using limma since two years. I have used ANOVA previously too but the problem was mainly with the traditional time course data. Because the data is longitudinal but they don't have any replicates. I was wondering how to do limma on this type of time course data. Anyways thanks

OK, so it would appear that doing ANOVA isn't actually the problem, in fact ANOVA doesn't seem to be applicable to your data.

If you edit your question to explain what the design of the experiment is, ideally by giving a detailed targets table, then we can suggest a more appropriate analysis for you. If the experiment consists of 3 timepoints for a number of patients, then a paired analysis is appropriate (see pairing or blocking in the limma User's Guide.) In that case, the patients represent the replication.

My experimental design and the target table looks as follows:

SUBJ.1720,SLE,baseline baseline Disease

SUBJ.1720,SLE,week16 week16 Disease

SUBJ.1720,week 52 week52 Disease

I don't have any healthy samples to compare. I don't have any replicates either. All I have is 400 samples at three different time points like the baseline, week 16 and week52. My goal here is to do differential gene expression analysis.So, I thought I could just do the ANOVA making three different comparisons like baseline vs week16, week16 vs week 52, and baseline vs week52. But I was wondering if this is the optimal way or is there any other good way to find the differentially expressed genes.

Thanks,

Prat

Dear Prat,

You haven't explained your experimental design in terms that I can understand. I cannot give you analysis advice from the limited information that you have provided.

With 400 samples this sounds like a serious data set, and there must be some appropriate analysis. One way anova however is definitely too simple. I strongly suggest that you consult a statistician or experienced statistical bioinformatician at your own university. Good luck.