Metanetx has a SPARQL endpoint:

https://rdf.metanetx.org/

Maybe it could be interesting to query it using UniProt protein ID to find reacitons?

Currently, EsMeCaTa will select a taxon if it contains at least 1 proteome.

An option to modify this number of proteomes could ask EsMeCaTa to select only taxon associated with at least X number of proteomes.

In __main__.py, version is assigned using an erroneous package name:

VERSION = pkg_resources.get_distribution("metage2metabo").version

A new version of UniProt is currently in beta. esmecata should be updated (maybe first with a --beta option).

New API available at: https://beta.uniprot.org/help/api

It seems that the following lines are not working:

    ?proteome up:organism ?organism .
    ?organism rdfs:subClassOf* taxon:{0} .

So it is not possible to retrieve the proteomes associated with a taxon, it is only possible to retrieve the proteomes directly associated to the corresponding organism.

And at this moment, the all esmecata proteomes step is not working with SPARQL.

With the failed 67f882f (and with local test), it seems that there is an issue with SSL. This could came from expired certificates associated to Mozilla trust store (as they are used by python-certifi).

During the proteomes or the annotation step, queries against UniProt can get stuck. Meaning that the tool process is still alive but no more proteomes/annotations are downloaded.

Need more investigation to understand why the queries get stuck:

• too many queries in a short time (a solution could be to increase time sleep)?
• downloading limitation on UniProt part?

So far all proteomes must have a protein in a cluster so that the cluster is kept for core proteome computation after MMseqs2.
It seems reasonable that a threshold could be used here to account for the incertainty and noise in the proteomes.
That threshold would be an option to esmecata clustering.