gget archs4 uses the legacy maayanlab.cloud endpoint; ARCHS4 has moved to archs4.org (evaluate migration, low priority)

[Elarwei001]

Background

ARCHS4 has moved to a new site and dataset at https://archs4.org (a banner on the old site, https://maayanlab.cloud/archs4/, announces this). gget archs4 currently queries the legacy endpoints (defined in gget/constants.py):

• tissue expression: https://maayanlab.cloud/archs4/search/loadExpressionTissue.php — POST, query ?search=<GENE>&species=<human|mouse>&type=tissue, returns CSV
• gene correlation: https://maayanlab.cloud/matrixapi/coltop — POST, JSON {"id": "<GENE>", "count": <n>}, returns JSON

Anyone can confirm the legacy tissue endpoint, e.g.:

curl -s -X POST 'https://maayanlab.cloud/archs4/search/loadExpressionTissue.php?search=STAT4&species=human&type=tissue' -H 'Content-Type: application/json' | head

Current status (not urgent)

• The legacy site "will stay available" per the banner, so gget archs4 still works today.
• A recent intermittent breakage on the tissue endpoint (the upstream CSV sometimes omitting the unused color column → KeyError) is already handled by dropping it with errors="ignore" (in the dev-CI repair PR fix(ci): repair archs4 + ELM live-data test failures #252).

What changed upstream

• The new ARCHS4 (archs4.org) is the most up-to-date version (≈922k human / ≈1.04M mouse samples).
• Its programmatic access is centered on the archs4py package / H5 files rather than the legacy PHP HTTP endpoints gget uses. Tissue output is still box-plot stats (min/Q1/median/Q3/max, log2(TPM+1)).

Concern

The legacy endpoints gget relies on may eventually freeze (stale data) or be deprecated.

Proposed (low priority)

Track and, when convenient, evaluate migrating gget archs4 to the new ARCHS4 (archs4py / new API). No immediate action required — the legacy endpoint still works and the color robustness fix keeps it stable.