question: can I construct a GEM from a list of KO identifiers?

[silvtal]

Thank you for developing the RAVEN toolbox

I have a list of KO identifiers from an already functionally annotated genome (used MicrobeAnnotator). My question is - seeing that this toolbox constructs GEMs from the KEGG databases, would it be possible to directly "feed" it the list of KOs?

[edkerk]

Interesting idea! There is no straightforward way to do this, but with a little bit of coding you could write a function that can do this.

To get started, you first want to generate the global KEGG model, containing all KEGG reactions:

keggModel = getModelFromKEGG();

Then you want to extract the KEGG Orthology annotations:

KOannot = extractMiriam(keggModel.rxnMiriams, 'kegg.orthology');

Let's remove all reactions without a KEGG Orthology annotation (from the keggModel and your KOannot list), as you would not have any matching genes for those reactions anyway:

noKO = cellfun(@isempty, KOannot);
keggModel = removeReactions(keggModel, noKO, true, true, true);
KOannot(noKO) = [];

Now comes the part where you need to write a custom function. What you basically want to do is go through the KOannot list and replace the KO numbers with the gene identifiers and convert this into gene associations (in model.grRules format). How easy this is depends on how unique the KO-gene matches are:

1. Does every gene have one KO and does every KO have only one gene?
2. Can a gene be annotated with multiple KOs?
3. Can a KO ben annotated to multiple genes identifiers?

With some combination of ismember, containers.Map or maybe some old-fashioned loops you can do this, but it very much depends on the layout of your input data. Do you know whether scenario 1, 2 and/or 3 is true for your data? Feel free to post a short sample of what the table looks like.

In the end, if you've replaced the KOs with the gene identifiers and formatted this as a model.grRules field, you can use standardizeGrRules to make the rxnGeneMat field.

[silvtal]

Thank you for your fast and very complete reply! It's a relief to know that it is indeed possible. I'm trying a different method, but if I can't make it work I will follow your detailed instructions. Thank you so much!!