Refinement of CH$COMPOUND_CLASS

[Treutler]

CH$COMPOUND_CLASS should be refined. My proposal is as follows:

• One substance can have multiple compound classes. E.g. the substance rutin can be, amongst others, a flavonoid, a disaccharide, and a quercetin O-glucoside. I suggest to allow any number of CH$COMPOUND_CLASS entries.
• The format of CH$COMPOUND_CLASS should be '[Database/Ontology] [ID/Class name]' (analog to CH%LINK). We should decide, whether we accept 'ID' or 'Class name' or both.
• The prefix 'Natural Product' or 'Non-Natural Product' should be obsolete.

E.g.: ChEBI CHEBI:28527 KEGG Flavonols ChemOnt CHEMONTID:0003531 ChemOnt Organic compounds / Phenylpropanoids and polyketides / Flavonoids / Flavonoid glycosides / Flavonoid O-glycosides / Flavonoid-3-O-glycosides

[schymane]

I agree that the Natural/Non-Natural product is not ideal. This is almost impossible to classify automatically and e.g. our environmental standards are certainly not all “non-natural” – hence we set all ours to “NA”. There are some good new methods around, that I have not yet tested extensively to see how these may work on our datasets. However, the definition of this field is set quite clearly in the record format – and the current specifications do allow multiple entries after this first field, see: 2.2.2 CH$COMPOUND_CLASS * ・ Category of Chemical Compound. Mandatory ・ Example CH$COMPOUND_CLASS: Natural Product; Carotenoid; Terpenoid; Lipid ・ Either Natural Product or Non-Natural Product should be precedes the other class names .

I feel we will either have to enhance the CH$COMPOUND_CLASS field in a way that is compatible with the current specifications, or consider introducing a new field, e.g. CH$ONTOLOGY or similar? @m-arita has a valid point regarding overwriting in #81 – here we will also risk overwriting original knowledge if we are not careful and this is dangerous.