PanApta complete registration evidence matrix

The 677 rows are registrations, not 677 independently verified molecular targets. EIF4A1 occurs twice with different source/species contexts.
Stable IDs derive from registry_row, never from names. Existing target/source context IDs and independently reviewed project-qualified SNCA branch links provide the joins.
Source mismatch links remain rejected. Fragment, complex, cell and domain material qualifications are preserved.
Exact molecular construct identity is not inferred from UniProt/reference identity.
Candidate, round, observation, read, file and project counts use distinct units. Null counts carry reasons; they are not zeros.
Source collection files may appear under related registrations. They must be deduplicated by run/file hash; their totals are never assigned to an individual factor.
390 processed aptamer archive run objects are provided as project context; no project-to-registry read total is inferred.
External author tables for RHAU53 and YTHDC2 domains remain distinct from public candidate integration.
The 25 published retained aptamer-selection target groups are reconciled by explicit target IDs. Unlinked groups remain visible in the separate table.
Scientific evidence support and HTTP/access verification are independent. Access checks describe only the actual tested response scope.
access_links counts registry-access.tsv entries, not unique URLs, experiments or verified downloads. An external archive page does not mean that PanApta serves the complete archive.
SNCA: S1/S2 are source-qualified selection branches with the same reported F-type fibril positive-selection material, not two verified molecular targets. Retained counts remain separate from the public derived case and complete internal archive counts.
The original 18 unmatched groups are individually explained in core-registry-reconciliation.tsv; two SNCA branches are now connected and 16 groups remain without a direct registration connection.
No sequences, raw reads, private paths, credentials, new source licenses or human approvals are included.

Classification correction 2026-10-23: two PRJEB70964 branches use source-qualified protein/RNA target filters. Original cell/DNA metadata, sequence records and all scientific counts are preserved. Source library chemistry is not a per-sequence chemical identity. See source-classifications.json/tsv.
Of the 16 groups without direct registry links, 12 are not applicable to a single-protein registration (cell/tissue/vesicle materials); four require distinct provenance, species or registry-coverage review. Not applicable is not a scientific approval or a claim of complete material identity. All 16 groups remain exported.

2026-10-24: mouse ST2 material resolution, HeLa simulation reconstruction and two registration proposals are linked at /web/source-evidence. Three groups retain specific follow-up work; the historical 677 registrations remain the published reference table.