Add stats for number of genes per MOD for which IBAs get dropped on the floor

[cmungall]

Discussed on managers call https://github.com/geneontology/pipeline/issues/300

A statistic that is very useful for GO is the number of genes that are not mapped to the reference proteome for which we are losing IBA annotations

An ad-hoc way to get this:

✗ curl -L -s ftp://ftp.pantherdb.org/downloads/paint/presubmission/gene_association.paint_zfin.gaf.gz | gzip -dc | grep -v ^ZFIN | cut -f2 | sort -u | wc
    3005    3018   28923
✗ curl -L -s ftp://ftp.pantherdb.org/downloads/paint/presubmission/gene_association.paint_mgi.gaf.gz | gzip -dc | grep -v ^MGI | cut -f2 | sort -u | wc
     200     213    1680

[overcounts because I am lazily not filtering comments etc but you get the point]

This is based on the assumption that the paint_MOD files default to uniprot for where there is mapping

This could be done more systematically as part of the pipeline, with stats files generated

I also think it might be nice to consider this as a go rule ("all genes with ancestral annotations should have unambiguous mappings to uniprot") such that these numbers could be shown in the general report dashboard, but this requires further discussion

[dustine32]

@cmungall I think this report has something close: https://docs.google.com/spreadsheets/d/1hMGJ8MFu1ozO3pHt44G2PPN9taWRPqEvb-1MK6WqVOI/edit#gid=1446471698

The section separated by taxon ID counts annotations rather than distinct genes but I added this section specifically to help spot large drops in IBAs from release to release.

If the annotation count isn't granular enough (need the distinct gene count) can I just add this new stat to this same report?

[cmungall]

I think the point is that we want something that is executed as part of the pipeline, with output in a standard place

[dustine32]

Ah, OK. There might be a good place in the go-stats code to add this. Like in the go-annotation-changes.tsv report.