Entering edit mode
7.8 years ago
seta
★
1.9k
Hi all,
I'm busy with doing expression analysis of an RNA-seq project from a non-model organism by edgeR software. I previously did de novo transcriptome assembly by Trinity. I performed GO analysis by agrigo for up- and down-regulated genes, but I found that several GO terms of interest overrepresented in the both up- and down-regulated gene list. Could you please tell me if it's normal? how can we interpret such a result? any comments and suggestion would be highly appreciated.
Best
Did you in your GO analysis take the background of the experiment in account? (i.e. did you limit the ' search genome' to what you could detect in your assay?) Which tissue are you working on and which are the overrepresented GO terms? Does that make sense that those are overrepresented for that tissue?
I am working on a plant that its genome has not been sequenced. So, I did de novo assembly transcriptome and annotated it against Arabidopsis proteome, the well-annotated model plant, and used the annotation results (TAIR ID) as a background. the overrepresented GO do not depend on the tissue, they are related to our treatment. Have you ever been experienced such thing in your work?
I've seen examples in which the tissue type dominates the enrichment analysis, since there was no correction for background (interesting read: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4561415/)
So it makes sense for your treatment? That's nice then :p But I can imagine that it's also biological sound. Say we have a hypothetical pathway with a group of kinases phosphorylating themselves and downstream targets as an activation, and the pathway also contains inhibitors/repressors of these kinases. In the case that the pathway gets activated the kinases could get upregulated and the inhibitors downregulated. This is obviously grossly oversimplified.
Thanks for your comment and paper. Yes, they make sense for treatment and also your explanation could have happened. However, It might be better to investigate the corresponding up and down-regulated genes within similar GO terms, but there are many genes and it's not possible to investigate them one-by-one. Please kindly tell me if you have any suggestions for it?
I don't have good recommendations for that, but your enrichment analysis should never be an end point of your project. Somehow you must connect this back to the biology and/or wet lab. It's an easy analysis but doesn't tell you a lot. Moreover, using multiple databases and multiple tools you will always find a method which tells you what you wanted to find. It's interesting to get a quick idea what pathways and processes might be involved in your treatment, but it doesn't tell the whole story. It's also virtually impossible to dissect based on your expression data which is a 'cause' and which is a 'downstream effect'. Perhaps you want to try to identify an upstream regulator responsible for the distortion your see in that pathway, but that's probably not straightforward. One of the first papers I found about it is http://www.nature.com/articles/srep12393, although I haven't read it and it's out of your field, it might give you some inspiration.
But the question is also what you aim to obtain with your analysis.