# ECOD develop291: the last PDB-only release (classification and database)

ECOD (Evolutionary Classification of protein Domains) classifies domains from
experimentally determined structures in the PDB into a hierarchy:
architecture, X-group (possible homology), H-group (homology), T-group
(topology) and F-group (family). develop291 (2024-04-12, PDB entries up to
2023-10-20) was the last release built from the PDB alone. Later releases
(develop292 onward, now the v29x series) also classify AlphaFold DB models.

> Schaeffer RD, Liao Y, Cheng H, Grishin NV (2017). ECOD: new developments in the
> evolutionary classification of domains. *Nucleic Acids Res* 45:D296–D302.
> https://doi.org/10.1093/nar/gkw1137

develop291 is also what the legacy website (`prodata.swmed.edu/ecod/`, PHP) served
until it was retired in 2026. This record is the archival copy of the release
files and of the database behind that site. The per-domain coordinates for every
domain are in a companion record (see Related works).

## Counts (database `info` table)

| | |
|---|---|
| Domains | 1,083,021 (21,196 manual representatives) |
| PDB entries / chains | 207,142 / 758,660 |
| F-groups / H-groups / X-groups | 12,948 / 3,717 / 2,459 |
| PDB included up to | 2023-10-20 |

## Files

| File | Contents |
|---|---|
| `ecod.develop291.domains.txt` | **Main list.** Every domain: uid, domain id, manual/automatic representative status, F-group id, PDB, chain, ranges (PDB and SEQRES numbering), UniProt accession, group names, assembly status, ligands. Comment header gives the columns. |
| `ecod.develop291.fasta.txt` | Domain sequences (1,082,930; header `uid\|domain id\|F-group id\|range`). |
| `ecod.develop291.F99/F70/F40.domains.txt` | Representative subsets at 99 / 70 / 40% sequence identity (116,385 / 67,809 / 44,289 domains), same columns. |
| `ecod.develop291.F99/F70/F40.fasta.txt` | Sequences of those subsets. |
| `ecod.develop291.F99/F70/F40.pdb.tar.gz` | Coordinates of those subsets, one `.pdbnum.pdb` per domain (115,737 / 67,466 / 44,129 files: a few domains in each list have no file). |
| `ecod.develop291.pdb.tar.gz` | Coordinates of the manual representatives (21,195 files). |
| `ecod_develop291.pgdump` | `pg_dump -Fc` of the website's PostgreSQL database (PostgreSQL 17), schemas `ecod`, `pdb`, `ecodf`, `pfam`, `public`. Restore with `pg_restore -d <db> ecod_develop291.pgdump`. |
| `ecod_develop291.tables.tar.zst` | Every base table of that database as TSV with a header row, named `<schema>.<table>.tsv`. |
| `MANIFEST.tsv` | Size and MD5 of every file. |

The release files are exactly as published at `prodata.swmed.edu/ecod/distributions/`
(2024-04-12). Paths inside the `.pdb.tar.gz` archives start with
`data/ecod/domain_data/<uid[2:7]>/<uid>/`. Unpack `.tar.zst` with `tar --zstd -xf`
(or `zstd -dc FILE | tar -x`).

### The database

The main table is `ecod.domain` (1,083,021 rows, the same uids as
`ecod.develop291.domains.txt`). The hierarchy is `ecod.cluster` and
`ecod.cluster_relation`. The other tables hold what the website showed alongside:
- PDB entry and chain metadata (`pdb.*`)
- taxonomy (`ecod.ncbi_taxa`, `ecod.ncbi_taxon_map`, `ecod.pdb_chain_taxa_map`)
- UniProt mapping (`ecod.unp_acc_map`)
- ECOD family (ECODf) accessions per F-group and their definitions (`ecod.ecod_pf_ids`, `ecodf.family`), and Pfam clans (`ecod.pfam_clan`)
- SCOP mapping (`ecod.scop_id_map`)
- DrugBank links (`ecod.ecod_drugbank*`)
- links from domains to human UniProt entries (`ecod.ecod_hs`) and SNV consequence terms (`ecod.dbsnv_disgen_consequence`)
- chain segments that are not classified as domains, such as expression tags, peptides, non-peptide polymers, coils and linkers (`ecod.special`, 138,068)

TSV notes: empty fields are missing values. Tabs and newlines inside text fields
are written as `\t` and `\n`. Array columns use PostgreSQL syntax (`{B}`).

`ecod.domain` and `ecod.develop291.domains.txt` agree on every uid, domain id,
PDB, range and T-group. They differ in how F-groups are written:
- **No F-group (59,029 domains):** the database has the T-group id plus `.0`
  (`1.1.1.0`); the release file has the bare T-group id.
- **665 domains with an F-group:** the release file has the bare T-group id but
  still gives the F-group name. The database has the full F-group id.

Five domains have an empty `seqid_range` in the database but not in the release file.

Website usage-statistics tables (`public.usage_rollup_*`) are not part of the
release and are not included.

## License and attribution

CC BY 4.0. Coordinates are derived from PDB entries (wwPDB, CC0).

When using these data please cite the NAR 2017 paper above.

Contact: Grishin lab, UT Southwestern Medical Center (http://prodata.swmed.edu/ecod/).