# Current VDJdb record links

The summer structure collection is preserved unchanged. This directory links it to the verified current VDJdb tables at `a911d825dfef5c9c3523892080fd2f83cc62b525`, using both submitted and repaired anchor-inclusive alpha/beta junctions and the peptide. Source structure metadata is at `f4efe2a278bebbd30eabf73c5328045798b437d4` in this dataset mirror. The metadata describes this exact database snapshot; publication of repository changes does not create a dated database release.

## Complete structure accounting

| Deposited PDB files | Predicted | Native | Total |
|---|---:|---:|---:|
| Exact paired junctions and peptide in current records | 8049 | 349 | 8398 |
| Historical repair or corrected source-observation link | 24 | 0 | 24 |
| PDB accession link with sequence or peptide differences | 0 | 7 | 7 |
| Discarded old 10X assignment | 2811 | 0 | 2811 |
| Excluded modified chemical ligand | 0 | 13 | 13 |
| Total files | 10884 | 369 | 11253 |

Thus 8429 deposited structures have at least one current record ID. All 15018 source metadata rows, all 11234 source hashes and all 55706 files across six archives are accounted for. Structure hashes and PDB accessions are different units: several native accessions share a sequence hash, and one hash occurs in both predicted and native metadata.

The 2811 unlinked predicted structures correspond to 2813 historical metadata rows. Every row matches the submitted alpha/beta/peptide key in the old `ISSUE_289.tsv` before the caller replacement (`c391ea9^`). Neither its modeled pair nor its recorded pre-repair pair occurs against that peptide in the current records. These are discarded assignments, not files lost during migration. Source changes are `c391ea9` and `e26bb90`; the existing caller audit remains the authority for the source decisions.

The 13 unlinked native accessions are `6v0y`, `6v13`, `6v15`, `6v18`, `6v19`, `6v1a`, `7byd`, `7ndt`, `8shi`, `8trl`, `8trq`, `8trr`, `9nig`. All were already explicitly excluded in `chunks_with_unconventional_aa/PDB_Database.txt`. Deposited citrulline, ABA, QM8 or lipid chemistry is missing or substituted in the old model peptide labels. They remain preserved with `disposition=excluded_modified_ligand`, literal chemistry and reasons in the native-review fields. They require a declared chemical-ligand representation before positive admission. [The native review](../native_review/README.md) gives the source-supported fate of all 14 investigated complexes. The valid standard-peptide `9GV6` observation has now been added as `VDJDB0000339155`; both modeled junctions and the peptide match its original and repaired states. Unreported V/J calls remain blank, so this observation has a current ID in the definitive tables and full export while the unchanged legacy chain/slim filter excludes it.

## Files

- `vdjdb_structures_metadata_current.tsv.gz`: every original column plus one row per source-metadata-row/current-record link. Unlinked rows are retained with an empty `record_id`. `source_row` is the 1-based data-row number in the preserved original metadata; `idx` remains the original legacy field.
- `record_links.tsv.gz`: flat link table, 160647 source-row/record links to 12788 distinct current record IDs. Repeated source annotations remain traceable through `source_row`; do not count link rows as independent structures or experiments.
- `source_disposition.tsv.gz`: exactly 15018 rows, one disposition/count summary per original metadata row.
- `asset_inventory.tsv.gz`: exactly 55706 archive members, their sizes and linkage dispositions. Filter `archive` to `pdb_files.tgz` or `pdb_files_native.tgz` for the 11253 deposited PDB structures.
- `asset_sources.tsv.gz`: flat mapping from `asset_row` to `source_row`; join through `record_links.tsv.gz` for asset-to-record links.
- `native_not_in_current_database.tsv.gz`: original metadata for the 13 excluded modified-ligand complexes.
- `verification.json`: exact input commits, SHA256 checksums, counts and source paths.
Regeneration and verification tools are in [`tools/`](../../tools/). Local execution provenance is retained outside published dataset files.

## Sequence states and association strength

`matches_original` requires both modeled junctions and the peptide to equal the submitted chunk record. `matches_repaired` requires both modeled junctions and the peptide to equal the current repaired record. Both true means repair did not change either modeled junction for that record. The separate `matches_original_junctions`, `matches_repaired_junctions` and `matches_epitope` fields expose sequence matches even when a native peptide construct differs. `current_*` columns retain the current IDs, sequences, pMHC context and chunk coordinates.

`sequence_match` counts over source-row/record links are:

| State | Links |
|---|---:|
| original+repaired | 160101 |
| original only | 3 |
| repaired only | 502 |
| neither, proven observation/accession lineage | 41 |

`link_basis` distinguishes the evidence:

- `paired_junctions_epitope`: literal equality of both junctions and peptide, across all current records, without a V-gene uniqueness filter.
- `historical_repair_lineage`: 8 source hashes link through the summer metadata's `cdr3fix.*.cdr3_old` pair and unchanged peptide to the current submitted pair. The modeled sequences differ from both current sequence states. This includes one retained 10X observation.
- `corrected_source_observation`: 25 source rows / 16 hashes link by publication, clone, subject and peptide to their surviving records. Of these, 23 records have the erroneous alpha chain removed and two retain corrected beta sequences. The model is not an exact model of the corrected pair. Alpha correction is documented in source commit `8077f14` / issue #681.
- `pdb_accession_provenance`: 7 native accessions link to 8 current records by the same PDB accession. These are `4nhu`, `7z50`, `8vq8`, `9aud`, `9k2r`, `9rup`, `6ulr`. Modeled peptide constructs or junctions differ from the current annotation; every difference remains visible.

## Limitations and use

A record-ID association does not imply that every modeled residue or pMHC label matches the record. Use `matches_original` or `matches_repaired` when a literal paired-junction/peptide match is required. Use `exact_tcr_pmhc_match` to require that match plus agreement on recorded species, MHC class and both MHC chains. For MHC comparison only, H-2/H2 spelling and HLA fields beyond the second field are normalized; expression suffixes remain distinct. No ambiguous allele or missing MHC call is inferred.

There are 104 source hashes with at least one literal paired-junction/peptide link whose recorded pMHC context differs: 25 predicted and 79 native hashes. A hash may also have a compatible link. Of the deposited structures, 8049 predicted and 292 native structures have at least one exact sequence/peptide/context link. The other links remain explicit for review; they do not silently transfer a model to another pMHC.

No fuzzy sequence matching, peptide trimming, donor collapse, source repair, structure modification or new scientific analysis was used. Original/repaired flags were independently checked against the output sequences, and all 12788 linked current record IDs were checked against the submitted junctions at their source-chunk coordinates. Every input checksum was verified after generation. Existing Polars joins sufficed; no parallelism or new dependency was needed.


## Low-confidence collection and complete PDB coverage

[The low-confidence collection](../../low_confidence/withdrawn_10x/README.md) contains the withdrawn 10X assignments, with `confidence_reason=withdrawn_assignment` and `experimental_negative=false`. These labels refer to assignment confidence, not model geometry. Of its 2,811 structures, 2,538 (90.3%) are assigned to KLGGALQAK. The other 273 include GILGFVFTL (117), RAKFKQLL (95), and six other peptides (61 combined). No hash matches a current positive paired-junction/peptide key, including other publications and both source/repaired sequence states. The paired receptors of 174 hashes do occur against other peptides; those positive observations remain unchanged.

[Coverage metadata](../native_review/pdb_coverage_current.tsv.gz) links all 360 declared PDB_Database accessions to current IDs and available coordinates. Eighteen accessions missing from the historical native archive were downloaded; complete deposited mmCIF files are under `data/pdb_sources/`. Those files retain their deposited frame and chemical residues. Seven pre-existing aggregate source rows have no declared accession and are listed separately; coverage is complete for every declared accession.


## Joining to the database

[VDJdb source](https://github.com/antigenomics/vdjdb-db) and
[build-output documentation](https://docs.isalgo.dev/vdjdb-db/outputs.html#structure-files-and-current-record-cross-links)
define the complementary database tables. Join `record_links.tsv.gz.record_id` to
`records.parquet.record_id`, then to `chains.parquet` on that ID. There is one chain row per
`(record_id, gene)` and potentially many source rows or structures per observation.
For coordinate members, join `asset_inventory.asset_row` to `asset_sources.asset_row`, then
`asset_sources.source_row` to `record_links.source_row`. Predicted filenames contain the source
`tcr_pmhc_hash`; native filenames contain the PDB accession. Use the accession as well as the hash
for native structures, because a hash can occur under several accessions.

These reviewed associations are broader than the database's automatic structure-evidence producer.
That producer applies additional V-gene and pMHC proof requirements and does not silently admit
the historical sequence-difference links listed here. Compare units and acceptance rules before
comparing counts. Download the primary database ZIP for Parquet and gzip TSV, the legacy ZIP for
existing positional text consumers, or the AIRR ZIP for AIRR tables; coordinates remain here.

## Current-record coordinate proofreading

The independent VDJdb proofreading script now checks PDB accessions across every current record, including observations retained under publication chunks after deduplication. It extends the earlier 360-accession aggregate-chunk check to 367 accessions across 411 current records. Six additional deposited files (`9NU7`, `9NV6`, `9NV2`, `9NVE`, `9NW2`, `9NVA`) are preserved under `data/pdb_sources/`; `7NA5` was already in the native archive. The script verifies all archived assets, source checksums and current record links, and writes separate coordinate-to-record and native-coverage Parquet tables. The existing historical archives and source-row associations are unchanged. Run `processing/proofread_structure_files.py` in the VDJdb source repository; see [the structural output contract](https://docs.isalgo.dev/vdjdb-db/outputs.html#structure-files-and-current-record-cross-links).

## Source-row refresh after duplicate correction

VDJdb issue [1369](https://github.com/antigenomics/vdjdb-db/issues/1369) removes redundant source rows. At database candidate `1e8415c5a6240d605b4ddf16380ba3766ddf27c6`, this shifts `current_chunk_row` on 27,976 link rows across 6,139 existing record IDs in both current link tables. Only that source-coordinate field changes. All other cells, structure associations, scientific context and sequence-match flags are byte-for-byte equal as decompressed values. No source-record links or coordinates are added or removed. The existing strict link audit passes against the corrected tables. Original reconciliation statistics and input identities above remain the history of the scientific mapping.