GO, slim, GO!

Comparing sample data in EBI Metagenomics is easier than ever, thanks to a new, specialised Gene Ontology 'slim' that creates a high-level visualisation of your sample's functional profile.

GO slims are cut-down versions of the GO that contain a subset of terms. Because they give a broad overview of GO content, rather than the full detail of fine-grained terms, slims are useful for summarising the processes or functions mediated by groups of proteins.

EBI Metagenomics uses a GO slim for high-level visualisation of the functional profile of metagenomic samples, which lets users run a quick comparison of samples. The original GO slim, developed in May 2012, was informed by several million GO annotations, but they were assigned to protein sequences from a very limited number of metagenomic studies. Because the slim was limited in focus and went out of date quickly with respect to the GO, it was time for a refactoring.

Rebuilding the GO slim

Thanks to interoperability work done under the auspices of ELIXIR, service teams at EMBL-EBI were able to rebuild the GO slim for metagenomics analyses. The new version is based on the analysis of 22 billion (22 x 10⁹) GO terms assigned to publicly available sample datasets analysed through EBI Metagenomics.

To be sure the GO terms in the pipeline were broad enough to describe the range of protein functions likely to be in a sample, the metagenomics team at EMBL-EBI evaluated each GO term and rated it according to the level of coverage it provides, then added new or alternative terms to improve specificity. When the list was ready, specialists in the ontologies team made sure the list was rich enough to catch as many annotations (i.e. descriptions of functions) as possible. To do this, they went through an iterative process, making sure annotations didn't fall through the cracks and going back to the selection process. This was repeated until the slim captured around 90% of annotations for all branches of the GO (molecular function, biological process and cellular component).

The EBI Metagenomics pipeline component that maps annotations to the slim was also completely rebuilt, using open-source tools (i.e. OWL). The metagenomics team deployed the improved slim and its software as part of a new EBI Metagenomics Analysis Pipeline version 3.0.

Future refinements

We hope the procedural knowledge and analysis scripts generated during this process will greatly facilitate any future rebuilding of, or refinements to, the GO slim. The GO slimming process will be repeated periodically to ensure the analysis pipeline keeps pace with changes to the GO. 

Feedback from our users is, as always, welcome and very much appreciated.