get metabolic pathways and metabolites into the KG

[saramsey]

No description provided.

[saramsey]

rationale: many diseases of interest to NCATS pertain to metabolic disorders.

[DeqingQu]

note from Steve： for starters we just need a TSV file with four columns, in which column 2 is the metabolite's KEGG ID (like KEGG:C12345) and column 3 is the human-readable name of the metabolite . so this file can be prepared any way you like, it doesn't need to come from RESTful querying of a web API

[saramsey]

Deqing, this REST API looks better than humancyc:
http://www.kegg.jp/kegg/rest/keggapi.html

[saramsey]

to get a list of all metabolites, you would do:

curl http://rest.kegg.jp/list/compound

(note that in the resulting TSV, instead of identifiers like "KEGG:C00022" you'll have "CPD:C00022"; this is fine).

to get a list of proteins that are associated a specific compound like KEGG:C00022, you would do:

curl http://rest.kegg.jp/link/ec/C00022

that will return TSV result like this:
cpd:C00022 ec:4.4.1.25
cpd:C00022 ec:4.4.1.28
cpd:C00022 ec:4.4.1.35
cpd:C00022 ec:4.4.1.36
cpd:C00022 ec:4.4.1.6
cpd:C00022 ec:4.4.1.8
cpd:C00022 ec:4.5.1.2
cpd:C00022 ec:4.3.3.7

the "ec:" identifiers are protein identifiers, but in a format called "Enzyme Commission". To convert from "ec:4.4.1.35" to "UniProt", we would do:

curl http://www.uniprot.org/uniprot/?query=(ec%3A+4.4.1.35)&format=tab&columns=id

Can you code this up please?

[saramsey]

So you will need to create a Python module QueryKEGG.py to expose a method "map_kegg_compound_to_enzyme_commision_ids"

You will then need to add a method "map_enzyme_commission_id_to_uniprot_ids" to the module QueryUniprot.py

[saramsey]

@dkoslicki OK we have a plan for this. I'm 95% sure we can get this done before the hackathon.

[dkoslicki]

Awesome!

[DeqingQu]

Thank you, @saramsey. Your instructions are very clear and I will finish the task ASAP.

I have two questions about the list of all metabolites.

1. The descriptions of some entries are quite long. For example,
   cpd:C00006 NADP+; NADP; Nicotinamide adenine dinucleotide phosphate; beta-Nicotinamide adenine dinucleotide phosphate; TPN; Triphosphopyridine nucleotide; beta-NADP+
   It seems that the description consists of many synonyms which are separated by semicolons. Should I only pick the first one before semicolons, or keep all of them?
2. Many of them are not related to human. Do I need to filter them? If so, is there any advice about how to do it?

[saramsey]

just to get this in the comment history: I like your idea of using the text up to the semicolon, as the "name". Then the complete text (including the text before and after the semicolon) can be the node "description".

[DeqingQu]

@saramsey The two methods in QueryKEGG.py and QueryUniprot.py are done and all test cases are passed.
BTW, I fixed a bug in the uniprot_id_to_reactome_pathways method in QueryUniprot.py module. 'Assertion Error: assert 200 == res.status_code'.

[saramsey]

[saramsey]