propagate counts considering part_of and other relationships

[dgpinheiro]

The same way as already mentioned by other colleagues, we are also interested in an implementation considering relationships, such as part_of, to propagate counts on them.

[dvklopfenstein]

I will take a look at this this week.

Thank you for taking your time to contact us regarding propagate counts using relationships.

[dvklopfenstein]

FYI: I just completed adding functionality for #126, which is a prerequisite for addressing your issue.

Withe #126 functionality implemented, I will now work on this issue.

Thank you very much for your interest in GOATOOLS, taking the time to write us, and for your patience. I look forward to implementing this issue.

[dvklopfenstein]

GOATOOLS now supports traversing optional relationships, like part_of and regulates, when propagating counts using these two new arguments in scripts/find_enrichment.py:

  -r, --relationship    Propagate counts up all relationships (default: False)
  --relationships [RELATIONSHIPS [RELATIONSHIPS ...]]
                        Propagate counts up user-specified relationships
                        (default: None)

You can specify all relationships using the -r argument. Or you can select specific relationships to use in propagate counts using the --relationships argument:

        # NO --relationships arg            -> None
        # --relationships part_of regulates -> relationships=['part_of', 'regulates']
        # --relationships=part_of           -> relationships=['part_of']
        # --relationships=part_of,regulates -> relationships=['part_of', 'regulates']
        # --relationships=part_of regulates -> NOT VALID

If you are running GOEAs from your own scripts, access this new functionality using the relationships keyword argument in an instance of GOEnrichmentStudy:

    goeaobj = GOEnrichmentStudy(
        geneids_population,
        assoc_geneid2gos,
        godag,
        propagate_counts=propagate_counts,
        relationships=VALUE,
        alpha=0.05,
        methods=['fdr_bh'])
    results = goeaobj.run_study(geneids_study, prt=prt)

Where VALUE is a set of relationships or a bool:

relationships=False  # Only required is_a relationships are traversed
relationships=True   # All optional relationships are used in propagate counts

# Traverse only the user-specified optional relationships:
relationships={'regulates', 'negatively_regulates', 'positively_regulates'}
relationships={'part_of'}

[dvklopfenstein]

@tanghaibao , Can you release a new version of GOATOOLS? All the tests are passing and this new version will include a substantial increase in functionality in using the optional GODAG relationships.

[tanghaibao]

@dvklopfenstein
Just upgraded to v0.9.7. let me know if there's an issue.

[Melnorme1984]

Maybe the update_association method here should be updated to use this new functionality: https://github.com/tanghaibao/goatools/blob/master/goatools/obo_parser.py

[dvklopfenstein]

@Melnorme1984 FYI: the update_association function in GODag will deprecated and has been replaced by the faster version in the GOEAs. The older version will be removed in a later release after deprecation.