KEGG annotation of ORFs -- rc4
======

KO source: eggNOG-mapper 2.1.12, eggNOG DB 5.0.2 (files dated 2020-2021,
i.e. a ~2019 KEGG vintage), `KEGG_ko` field.

 - orf-to-ko.map.xz: Mapping of ORFs to KOs, from ../../rc4_functional/
   orf_to_ko.tsv with the `ko:` prefix stripped, de-duplicated, sorted.
   Distinct KOs observed: 17,108.

Reference hierarchy: regenerated fresh from the live KEGG REST API
(rest.kegg.jp), NOT copied from WoL2's kegg/ (that is a frozen 2022
snapshot covering only 10,237 KOs, against the 17,108 observed here). Two
real KEGG REST API quirks were found and worked around while building it:

 - `/link/X/Y` returns rows in the *reverse* column order the URL naively
   suggests (verified against three independent endpoint pairs).
 - Only KO's own pathway linkage exposes both the species-generic
   "map#####" and ortholog-view "path:ko#####" forms; every other entity
   linked to pathway (module, compound, disease, reaction) returns only the
   "map#####" form. Handled by converting map->ko by shared numeric ID
   rather than filtering (filtering produced empty files on the first
   attempt -- caught before shipping).
 - pathway-to-disease.map: KEGG's own `/link/disease/pathway` only links
   disease to organism-specific pathways (e.g. path:hsa05340), never the
   generic/ortholog form, so this file is instead derived transitively via
   pathway-to-ko join ko-to-disease (matching the parent README's
   collapsing-order diagram, where disease hangs off KO, not off pathway).

ko-to-go.map: KO-keyed, NOT the same file as ../go/idmaps/kegg.map
(reaction-keyed). KEGG REST has no `go` linked db for `ko` (`/link/go/ko`
-> HTTP 400), and GO Consortium's external2go has no KO-keyed KEGG
cross-reference either (only kegg_reaction2go, which is what
../go/idmaps/kegg.map is built from -- see ../go/README). So this file is
built from eggnog.db's `prots` table instead (kegg_ko + gos columns,
285,616 reference proteins with both populated) -- an aggregated empirical
cross-reference from eggNOG's reference protein set, NOT a single curated
per-KO definition. Spot-checked against ec2go for
consistency: K00001 (EC 1.1.1.1, alcohol dehydrogenase) correctly resolves
to GO:0004022, matching ec2go's own EC:1.1.1.1 entry.

ko_name.txt backfill: 23 of our observed KOs are absent from current KEGG
(retired/renumbered since eggNOG 5.0's ~2019 vintage) but are still present
in WoL2's own frozen 2022 snapshot (KEGG dropped them sometime between 2022
and now) -- backfilled from
/projects/wol/qiyun/wol2/function/kegg/ko_name.txt. These 23 rows reflect
KEGG's 2022 state, not current KEGG, for a KO current KEGG no longer defines
at all -- listed separately in ko_name_backfilled.txt for transparency; not
distinguishable from the other 28,402 rows in ko_name.txt itself. The
remaining 105 aren't recoverable from anything available locally (KEGG's REST
API has no historical/versioned endpoint).

Regeneration and KEGG release: KEGG's REST API always serves the current
release, so a later rebuild will pick up whatever KEGG has published since.

Format: see ../README's "Format contract" section. module-to-reaction.map
and pathway-to-module.map are WIDE (needed by `woltka coverage`); every
other file here is long/strict-2-column.