Domain structures for 4.9M non-singleton representatives of a provisional classification of the AlphaFold database 200M set
Year: 2026
Type: dataset
Abstract: The predicted structural archive behind A topology census of four-strand β-sheets in experimental and predicted protein structures: 4,959,302 DPAM-parsed domain structures for the non-singleton representatives of a provisional classification of the AlphaFold database 200M set. Why these are deposited rather than cited. The AFDB 200M classification they come from has not been published, so there is no archive to point at. This record is the only source. Full-atom PDB, packed as 20 xz archives of roughly 1.6 GB, each covering five shards of the source store. Members are named {ACC}_D{n}.pdb or dpam_{ACC}_nD{n}.pdb -- both spellings occur; an index built on only the dpam_ form returns 1.22% of the store, and one built on only the bare form silently drops that same 1.22%. The shard a domain falls in is not derivable from its accession; to find one domain, look across all shards. README.md states what the record is and how its counts reconcile; MANIFEST.md and SHA256SUMS give per-archive checksums with source-tarball and member counts, each verified against an independent count read from the source tarballs rather than from these archives. This is a superset of what the census searched, by 0.91%. The paper reports 4,914,196 representatives, which is meta-matrix records minus 7,735 seeded ECOD references. The 45,106 difference is domains whose structures exist here but which produced no meta-matrix record — 44,978 in the new/ tree and 128 in dpam/. The store is the classification's structures, not a transcript of the census input. The census itself — code, artifacts and the per-topology counts — is the companion record 10.5281/zenodo.21814885. The 7,746 ECOD reference domains seeded into that clustering are not included: they are experimental structures, and the paper counts 7,735 of them as experimental rather than predicted.
Institution: The University of Texas Southwestern Medical Center
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7204113544