Binning GO terms into broader (more usefull) GO terms

Question

[T_18]

Dear all,

For quit some hours I have been looking for a way to bin a list of GO terms into broader, more functional, (GO term groups). More specific: I have a list of 100s of GO terms with quit specific information and I want to bin/group them into useful groups like: "development" and "metabolition", etc..

The end goal is to get more functional groups of my total list of differentially expressed genes.

E.g. from GO:0006694 to the broader group: GO:0008152 (metabolition process).

I'm sure this is not something new.. Closest I get is cateGOrizer: https://www.animalgenome.org/cgi-bin/util/gotreei But this means I need to do a lot by hand.

THanks in advance!

[T_18]

Thanks for your reply. To be precise about my goal: I have a list of 2500 differentially expressed annotated transcripts with GO terms. I think it is very usefull to split these 2500 genes into smaller chunks according to 'function' based on the GO terms. These can be relatively big like, development, metabolition, digestion etc.. But the GO terms I have now are super specific and I do not want to go through them by hand. Therefor I have been looking for a way to bin the GO terms into broader functional groups so I am able to devide my genes into smaller chunks.

Unfortunately I have not been able to find a simple solution, but I am sure I am not the first in this. I am thinking of something like: https://www.yeastgenome.org/cgi-bin/GO/goSlimMapper.pl But than with GO terms as input (that are than divided into similar functioning-groups) instead of gene names, and off course not for yeast in specific.

[Jean-Karim Heriche]

As I suggest below, you could just define a list of GO terms of interests then assign your genes to these terms using the current annotations you have. Note that this is equivalent to building a flat GO slim.
If you don't want genes to be in multiple categories, define an order for your terms and assign each gene to the highest ranking term it maps to.
If you want to use a GO slim, you first need to create or obtain a slim ontology file in obo format then you simply run the map2slim.pl script or owltools with the --map2slim option.

Answer 1

[Jean-Karim Heriche]

As usual, there are multiple approaches and what is most suitable depends on the context and what you're trying to achieve with this. One way is simply to define a list of terms that you're interested in plus a catch-all category called other processes. This way you get to choose the most relevant semantic level for each process. Alternatively, you could use a GO slim, i.e. a customized subset of GO, see this GO subset guide on the GO website. For an automatic way, you may be interested in this paper: Finding New Order in Biological Functions from the Network Structure of Gene Annotations. Also if you're interested in the link between function and phenotype using ontologies, you may be interested in this paper: How can functional annotations be derived from profiles of phenotypic annotations? which uses the idea from the paper mentioned above.

Answer 2

[Benn]

I know the feeling, the one of trying to figure out how to make GO terms more useful (and publishable) than just a list of terms. I decided that I wanted to cluster GO terms based on overlapping genes and make it into a heatmap, so you can publish a figure (instead of a long supplementary table). I made functions in R, which can be found in the package gogadget, see here.

I don't say my approach is the best or you should use it, also my package is meant for RNAseq expression data (enrichment with goseq, then cluster the GO terms with hierarchical clustering). Maybe the clustering approach can be implemented to your case, maybe not. Good luck with it.

[T_18]

This seems indeed exactly what I was looking for! I'm currently following the analysis steps, but I am stuck. (That is not part of this question so do not want to put that here).

[Benn]

Nice, I hope it can be of help. If you get stuck open a new question here on biostars with the gogadget tag. I will try to find a solution.