Metadata store follow-ups from first release (cmgd_mpa4.2-2.3.0 / 2026-10-16)

Found while publishing the first release (epic #266):

age_years normalization. cmd_sample_values and the harmonized view mix "23" with "23.0", and they contain "Not applicable" (seen in LassalleF_2017). Normalize numeric fields at load time and treat cMD's NA tokens as missing, as compare_cmd_olm already does.
Re-run nf-meta load xref now that the 470 MVP readsets are registered in v2. Their xref rows were predicted from Olm's and cMD's run groupings, and need confirming against Phase 3's ENA grouping.
LassalleF_2017 has no BioSample in samples because it isn't in Olm. Take biosample from v2 sample records or ENA instead of from Olm.
Delete the private HF test repo seandavis/cmgd-mirror-test once mirroring is signed off.

Releases are immutable, so these fixes land in the next release, the one that includes the 470-sample cohort.

🤖 Generated with Claude Code

Navigation Menu