GVClass reference resources v1.2.1

Reference database archive for GVClass v2.1.0. What is new in v2.1.0 Prokaryotic proteins removed from eukaryotic reference genomes: 2,246 reference proteins (3,083 marker rows). Two sources named them: the metaSAGs single-cell screen (near-identical placements of confident prokaryotic genomes on eukaryotic reference proteins) and a DIAMOND screen of the eukaryotic reference proteins of this bundle against its bacterial and archaeal reference proteins (a hit at 90% identity or more and no eukaryotic homolog at 70% identity or more in another genome). Rows by source: {"metasags_handoff_20260930": 3083}. The removed proteins and their evidence are listed in provenance/v2.1.0/removed_reference_sequences.tsv. No reference genome was removed. RESOURCE_MANIFEST.json lists the SHA-256 digest of every member. The bundled completeness and contamination models, the HMMs, the labels and the marker tables are unchanged from v2.0.0. Marker conservation tables and completeness baselines were not recomputed after the removal; 464 genome-marker pairs lost their only reference protein. The archive is stored as 24 parts of 64 MiB (resources_v2_1_0.tar.gz.part000 to part023) with two checksum files; GVClass v2.1.0 or later downloads, checks and joins them. To join by hand: cat resources_v2_1_0.tar.gz.part[0-9]* > resources_v2_1_0.tar.gz and compare with the SHA-256 in resources_v2_1_0.tar.gz.sha256. Use this archive with GVClass v2.1.0 or later.