BGC Atlas — BiG-SLiCE clustering runs (clustering v3), published 2026-09-21

The two BiG-SLiCE 2.0.2 runs the site's gene cluster families were built from,
each published whole as a single `result/data.db`.

bigslice-v3-seed.tar.gz   13.6 GB
  extracts to  bigslice-output-v3/result/data.db  (31.3 GB)
  sha256       5e75c403be1cb3a6863a25ac225f2925a4eb93301bdf6dad03189ce04291becf
  the main atlas families, clustered from the dedup-95 seed
    clustering 1: 371,651 families at threshold 0.4

bigslice-v3-residual.tar.gz   6.1 GB
  extracts to  bigslice-output-residual-v3/result/data.db  (14.5 GB)
  sha256       874757a92311362b2fb01ff5c051502546f857615e35bb6a48779fbf94cb7226
  de-novo families of what the seed run could not place
    clustering 1: 394,186 families at threshold 0.4
    clustering 2: 107,451 families at threshold 0.7

FAMILY IDS
----------
`gcf.id` in each file maps to a family on the site:

  seed run      GCF-{id:06d}           e.g. gcf.id 336 -> GCF-000336
  residual run  GCF-F-{400000+id:06d}  e.g. gcf.id 42  -> GCF-F-400042

The two are INDEPENDENT id spaces. gcf.id 42 in one is unrelated to gcf.id 42
in the other. Residual clustering 2 (@0.7) was never loaded and has no pages.

`bgc.name` is `genomes/<uuid>.regionNNN`; strip the prefix and suffix to get
`bgcs.id`, the key the other downloads use.

VERIFY
------
  sha256sum -c SHA256SUMS