fix: clean-up annotations

[edkerk]

Description of the issue:

Some overlap with #33 and #44, but bigger scope here. If this issue is settled, then the other two can also be closed.

Various (minor) issues exist with current annotations of genes, mets and rxns:

• [x] 1. metanetx is sometimes misspelled as metabetx

• [x] 2. About 21 metabolites are annotated to 3dmet. Either annotate all metabolites (as far as possible) to this database, or drop the annotation of these few metabolites. This is not a database commonly used for GEMs, the model is not using 3D structures of metabolites, so drop the annotation.

• [x] 3. About 19 metabolites are annotated to cas. Either annotate all metabolites (as far as possible) to this database, or drop the annotation of these few metabolites. This is not a database commonly used for GEMs, so drop the annotation.

• [x] 4. About 27 metabolites are annotated to pubchem.substance. The more common database is pubchem.compound. Drop the pubchem.substance annotation and make sure that the pubchem.compound is included for those metabolites.

• [x] 5. Only N-Acetyl-L-Cysteine is annotated to kegg.drugs, while kegg.compound is also defined. Drop the kegg.drugs annotation.

• [x] 6. Some ChEBI annotations are misrepresented, as CHEBI%3A15377 instead of CHEBI:15377. Each of the metabolites is also annotated with the correct ChEBI annotation: remove the incorrect ones.

• [x] 7. About 17 reactions are annotated to kegg.ontology. This is mostly redundant to the more commonly used kegg.reactions, which are annotated to 1369 reactions (including the 17 reactiosn with kegg.ontology). Drop the annotation.

• [ ] 8. Of the 1778 genes, 215 are annotated with GO terms. Should this be extended to all genes, or should go as annotation be dropped? See #44

• [ ] 9. Of the 1778 genes, 219 are annotated with PFAM domains. Should this be extended to all genes, or should pfam as annotation be dropped? See #44

Expected feature/value/output:

• Do not include sporious annotations to databases that are not commonly used:
  • Either remove the uncommon database, ...
  • ... or increase the coverage to all members.
• Improve coverage of common annotations (e.g. SEED, currently covering 1.9% of metabolites), will be handled in a separate issue.

To do:

If agreed to remove the sporious annotations to uncommon databases, to do list will be included here.

I hereby confirm that:

• [x] This problem persists in the master branch of the repository
• [x] A similar issue does not already exist
• [x] I've considered asking this first in the Gitter chat room

[edkerk]

Additional realisation: what is currently in the BioCyc field are actually MetaCyc identifiers. E.g. memote (see example report here) checks for BioCyc identifiers. To turn MetaCyc identifiers into BioCyc identifiers they have to be prefixed with META::

• META:F16ALDOLASE-RXN instead of F16ALDOLASE-RXN

These identifiers are then also following specifications on Identifiers.org.

• [ ] Correct BioCyc identifiers by adding META: prefix.

[sulheim]

• [ ] Fix annotations issues raised by @smoretti in #111

[edkerk]

The GO term and PFAM annotations to the genes will first be discarded (PR #117), as these are assigned inconsistently (only about 200 of 1778 genes have these annotations). Meanwhile #44 remains open, where additional gene annotations are proposed.

[sulheim]

Actually, we shouldn't discard the gene annotations. I haven't checked the 200 genes with annotations, but most genes are annotated to GO/PFAM/ETC in iAA1259. These annotations are valuable and should not be discarded.

[sulheim]

These annotations are already in this repository under data/sulheim2020/annotations/genes.csv. We can add GO and PFAM annotations from this spreadsheet into the model file.

[edkerk]

Ah, we already have that data, great. This will then be done in #44 and not addressed here.