I need to work on a published RNAseq cancer dataset (normal, primary and metastatic tumor). It's a pair end experiment. The published article only investigated the differential expressed genes, their association with the disease aggressiveness and they did GSEA.
I'm very interested in the disease and I was thinking of re-analyzing the data while exploring the disease pathogenesis from a different point of view. I thought I might go and do some network visualization and analysis of the gene expression data (in their publication they didn't do that). Also, studying the alternative splicing patterns and their involvement in different biological processes and cancer etiology.
I'm doing this for my Master thesis. My questions are:
Is there any other aspects that I should investigate using this data?
In case of using published data, shall I have a permission from the study PI or something to use their data?
I analyzed single end but I didn't do paired end. So, while mapping using Tophat, what would be the command look like in case of paired end?
Without knowing the details of the original study I would say that you're covering the major bases.
No, you do not need permission from the original PI, at least as long as they've made things publically available. Having said that, it's often nice to reach out to her/him and ask if someone is already working on this or if they might like to collaborate. In the former case this let's you know if you might be competing with someone (who's likely further along on things). In the latter case, you might be able to get some advice regarding the things they didn't publish, like observations about particular patients/samples that might need to be treated specially for some reason. Many things like this don't end up getting written about and being able to just quickly ask the group that produced the data can save considerable time.
Instead of tophat2 options genome file.fq, you'd use tophat2 options genome file_1.fq file_2.fq. You'll need to use both files like this to get any decent alternative splicing results.
1) This is for you and your supervisor to determine, that's the best part of grad school, figuring out what you want to study and look at :) You should be clear in terms of what you mean by analysing the gene expression data and saying the group didn't do that, only looked at differential expression. In gene expression experiments we are always looking at gene expression differences. Case versus control, condition one versus condition two, tumour versus matched normal, tumour versus tumour, etc. Alternative splicing is a good place to start as you mentioned. If they haven't looked at that it could be a gold mine of interesting data. Looking for fusion transcripts and like could also be interesting.
2) If the data has been published and deposited in a public repository than chances are no you don't. If it comes from a large consortium like the TCGA, you might want to check because data becomes publicly available in advance of any publications but there are often reserved analyses. So even though they have one pub out other analyses you want to do might be reserved. The other thing to be aware of here in terms of your own ability to publish is if this group IS doing some additional analyses on the dataset as part of a broader research programme, you may very well get beaten to the punch. They could already have a paper in the works doing a lot of the analyses you want to do. That is definitely something for you and your supervisor to think about and consider.
3) It should be in the manual. You want to become VERY familiar with all software you intend on using in your research to ensure you are using correct/reasonable parameter values for instance.
Thank you so much!