[iskandr]

Closes #204. Bumps to 0.51.0.

varcode will infer T2T-CHM13v2.0 from file names and VCF headers (openvax/varcode#723). For that it needs one list of names, contig aliases and mitochondria that both packages share. Until now ORFdata had only a private, hint-only alias dict, and load_genome accepted only "t2t-chm13".

Change

The table lives beside the pin in orfdata/data/genomes.json, so a second pinned assembly is a registry entry rather than new code.

• orfdata.reference_names() (orfdata genomes names) returns a DataFrame with columns name, genome, mitochondrion and note:

Name	Genome	Mitochondrion
t2t-chm13, T2T-CHM13v2.0, GCA_009914755.4, hs1, chm13v2.0, chm13v2.0_maskedY, chm13v2.0_noY	t2t-chm13	CP068254.1 (CHM13's own)
chm13v2.0_maskedY_rCRS, chm13v2.0_maskedY.rCRS	t2t-chm13	NC_012920.1 (the rCRS)
GCF_009914755.1	t2t-chm13	none (RefSeq gives MT no accession)
T2T-CHM13v1.1, GCA_009914755.3, chm13.draft_v1.1	t2t-chm13	CP068254.1
T2T-CHM13v1.0, GCA_009914755.2, chm13.draft_v1.0, chm13.draft_v0.9, T2T-CHM13v0.7, GCA_009914755.1, chm13.draft_v0.7	none, with the reason

• orfdata.contig_aliases(name) (orfdata genomes contigs NAME) lists the 25 sequences from NCBI's assembly report. Each row has the annotation's name (1), the UCSC name (chr1), the GenBank and RefSeq accessions, and the length. It leaves out sequences that the named files lack, or carry as a different molecule:
  • no Y for _noY and v1.1;
  • no MT for the _rCRS set, whose chrM is the rCRS;
  • no MT for GCF_009914755.1.
• load_genome, install_genome and orfdata genomes install accept any listed name, ignoring case. NotInstalledError names orfdata genomes install t2t-chm13 whichever name was given: "The t2t-chm13 annotation (for hs1) is not installed; …".
• catalog._assembly_hint takes T2T names from the table, so bare chm13 (which names every version) no longer resolves. A v1.0 name explains why it doesn't resolve.

Finding names inside paths stays with the caller; the table holds exact names only.

Sources

Checked 2026-11-09:

• NCBI's assembly reports for GCA_009914755.4, .3 and .2, plus NCBI Datasets for the assembly names (.1 is v0.7, .2 is v1.0, .3 is v1.1).
  • v1.1 has the same 1–22, X and MT accessions and lengths as v2.0, and no Y.
  • In v1.0, 23 of 24 lengths differ (chr1 is 248,387,497 bp vs 248,387,328). The marbl/CHM13 release notes say v1.1 filled the rDNA gaps and re-polished telomeres.
• UCSC's hs1.chromAlias.txt, which agrees with NCBI's report on every row.
• The marbl/CHM13 README: _noY is "identical to T2T-CHM13v1.1", and _maskedY_rCRS replaces the mitochondrion with NC_012920.1.

The two reports and the chromAlias file are in tests/data/t2t/, so the contig table and the v1.0 claim are tested offline against their sources. The report's SHA-256 is pinned beside the table.

Not changed

• The pinned Ensembl files and their cache digest, so existing t2t-chm13 installations are reused.
• No mitochondrial annotation yet: Ensembl's rapid-release T2T GTF has no MT genes (Annotate mitochondrial variants in T2T-CHM13 VCFs varcode#722).

Tests

tests/test_genomes.py covers:

• every name in the table resolves to its genome and mitochondrion, ignoring case;
• older versions raise their reason from _pinned, contig_aliases and load_genome, and bare chm13 doesn't resolve;
• the contig table matches NCBI's report and UCSC's aliases;
• each name lists only its own contigs;
• the not-installed message for an alias;
• an installed slice genome opens under its aliases;
• the catalogue hints;
• the new CLI commands.

tests/test_cli_audit.py runs genomes names and genomes contigs with human and JSON output.

Verification

• ./lint.sh: passes.
• pytest tests -n 8: 1527 passed, 6 skipped.
• mkdocs build --strict: passes.

https://claude.ai/code/session_01JHJhwfx4um52Nczrv3G814