What comes after cuffdiff?

[shurjo]

I guess this is a pretty open-ended question, but maybe the community will have some interesting discussions on it, so I'm throwing it out all the same. The dilemma I'm facing: I ran the Tophat/Cufflinks pipeline all the way to cuffdiff on the case and control groups in my project (eight subjects (replicates?) in each group). The RNA source we used to compare the two groups is not likely to contain large, consistent and dramatic changes between the groups, rather maybe very subtle changes. All of the .diff files from cuffdiff have numbers of transcripts ranging from 100-200 (out of 20000 RefSeq transcripts) that are significant as per cuffdiff's tests for expression/splicing/regulation. Now, what comes next? There are a few obvious things: qPCR validation of the best candidates, pathway analysis. The other approach might be to report results from cufflinks/cuffdiff alone as findings in their own right. Any thoughts from the larger universe of transcriptome-miners out there?

[severin]

Barely scratching the surface

You are just getting warmed up once you get through CuffDiff. Though I must admit it depends on what kind of biological questions you are asking with your experiment. Sure, metabolic pathways or a Gene Ontology or Kegg or some other form of annotation is one analysis that is performed. You can use your next generation sequencing data to improve annotation, verify assembly. Although annotation and assembly are important, many people don't find that to be very interesting. You will also likely find that many of your genes (depending on your organism) are not annotated or the annotation is based on the closest model organism. Gene expression atlases of multiple tissues can give you insight into function for some of your differentially expressed genes that are not well annotated. Then of course you could also explore the promoter regions and look for cis-regulatory elements, you could explore Transposable Elements, and the good old fall back of insertions and deletions and SNPs. However, if you are interested in evolution, well that adds a whole other level of interesting questions. Are any of these genes that are differentially expressed part of a family of genes and how did this family of genes arise? Was it due to large scale duplication/triplication events that are found in the majority of the eudicots or a result of several local duplications. Unfortunately, your question might be a bit too open ended. What kind of "case" and controls do you have? How did you set up your experiment? There is so much more that can be done but I will stop here. I love science.