Document database bundles and bidirectional structure links

README.md

@@ -19,7 +19,7 @@ pretty_name: vdjdb_structures
 
 This repository contains structure data for selected VDJdb records obtained using AI-based modelling in `data/` folder:
 * `pdb_files.tgz` contains predicted TCR:pMHC structures with canonical chain names, orientation and placement, superimposed by aligning and rotation. File names start with `tcr_pmhc_hash` which must be used for connecting the structure with the VDJdb record.
-* `pdb_files_native.tgz` contains real TCR:pMHC structures from PDB, processed to canonical coordinated in the same way as `pdb_files.tgz`.
+* `pdb_files_native.tgz` contains experimental TCR:pMHC structures from PDB, processed to canonical coordinates in the same way as `pdb_files.tgz`.
 * `coordinates_aa.tgz` contains tables with amino acid Ca atom coordinates for each residue of deposited structures.
 * `contacts_aa.tgz` contains files with residue pairs at a distance <= 5A for each of the deposited structures. Currently only TCR alpha CDR3-peptide and beta CDR3-peptide are considered.
 * `complementarity_maps.tgz` and `complementarity_maps_simplified.tgz` contains 2D projections of TCR alpha CDR3, TCR beta CDR3 and peptide residue coordinates in for of SVG plots produced by matplotlib. 
@@ -42,13 +42,13 @@ Metadata for all structures is concatenated with corresponding VDJdb record meta
 | MHC class | 14,915 MHC-I, 103 MHC-II |
 | species | 14,928 human, 90 mouse |
 
-A metadata row without a matching PDB is normal — the file set is the subset that modelled and passed
+A metadata row without a matching PDB is normal - the file set is the subset that modelled and passed
 QC. Join on `tcr_pmhc_hash`; predicted filenames are `aligned_aligned_<hash>.pdb`, natives are
 `aligned_<pdbid>.pdb`.
 
 ## Provenance
 
-- **TCR:epitope pairings are EXPERIMENTAL** — VDJdb records, carried through with the full VDJdb
+- **TCR:epitope pairings are EXPERIMENTAL** - VDJdb records, carried through with the full VDJdb
   "full table" columns (`reference.id`, `method.*`, `meta.*`, `vdjdb.score`).
 - **Predicted structures are COMPUTED** (TCRmodel2/AlphaFold). `ranking_confidence`, `plddt`, `ptm`,
   `iptm`, `tcr_pmhc_iptm` are the generator's own confidence metrics, not measurements.
@@ -60,8 +60,8 @@ QC. Join on `tcr_pmhc_hash`; predicted filenames are `aligned_aligned_<hash>.pdb
 ## Used by
 
 The TCRen2 receptor-ranking benchmark (`isalgo/tcren_structures`) draws its **real-binder positives**
-from this pool — filtered to `is_native == False`, restricted to hashes present in `pdb_files.tgz`,
-and with TCRvdb pairs excluded — then matches them per epitope against the decoys in
+from this pool - filtered to `is_native == False`, restricted to hashes present in `pdb_files.tgz`,
+and with TCRvdb pairs excluded - then matches them per epitope against the decoys in
 `isalgo/tcren_structures::vdjdb_negatives/` and `::immrep23_negatives/`. See
 `isalgo/tcren_structures::vdjdb_binder_benchmark/README.md`.
 
@@ -76,3 +76,46 @@ Of the 2811 structures in the low-confidence withdrawn-10X collection, 2538 (90.
 ## Native source review and PDB coverage
 
 [The native review](metadata/native_review/README.md) accounts for all 14 previously unlinked native complexes: 9GV6 now has current record ID `VDJDB0000339155`; the other 13 were already excluded chemical-ligand observations and remain preserved with their literal chemistry. [Coverage metadata](metadata/native_review/pdb_coverage_current.tsv.gz) confirms coordinates for all 360 declared PDB_Database accessions. The 18 missing accessions were downloaded; complete mmCIF sources, including the 14 reviewed complexes, are in [data/pdb_sources](data/pdb_sources/README.md). Current metadata retains original/repaired match flags and explicit mismatch/exclusion reasons.
+
+
+## Database builds, identifiers and downloads
+
+The curation source and build are [antigenomics/vdjdb-db](https://github.com/antigenomics/vdjdb-db).
+[Build-output documentation](https://docs.isalgo.dev/vdjdb-db/outputs.html) describes all three
+database bundles, their units and filters, compact Parquet/gzip tables, and the structure joins.
+Coordinates remain in this dataset; database bundles contain evidence IDs and supplied metrics.
+A database build does not copy these archives into every format.
+
+Use `metadata/current/record_links.tsv.gz` to join to a database build's `records.parquet` on
+`record_id`; join `chains.parquet` on the same ID for TRA/TRB sequences. The exact database input
+revision and SHA256 values are in `metadata/current/verification.json`. Read
+[the current metadata contract](metadata/current/README.md) before filtering original/repaired
+matches or resolving many-to-many archive links. `idx`, `complex.id` and the historical source
+hash are not substitutes for the current `record_id`.
+
+| Location | Purpose and format |
+|---|---|
+| `vdjdb_structures_metadata.tsv.gz` | Preserved summer-release source metadata |
+| `data/*.tgz` | Preserved canonical PDBs, residue coordinates, contacts and SVG maps |
+| `metadata/current/` | Gzip TSV links, source/asset dispositions and verification manifest |
+| `metadata/native_review/` | Fourteen reviewed native fates and coverage of declared PDB accessions |
+| `data/pdb_sources/` | Supplemental complete deposited mmCIF sources, gzip-compressed, with hashes |
+| `low_confidence/withdrawn_10x/` | Withdrawn-assignment metadata and byte-verified selected assets |
+| `tools/` | Regeneration and verification commands; see each script's `--help` |
+
+The low-confidence archives deliberately retain selected files also present in the historical
+mixed archives, providing a usable standalone partition while preserving the original collection.
+Download only the archive types needed for an analysis. Current links and dispositions are flat
+gzip TSV rather than uncompressed or repeatedly joined text copies. The expanded current metadata
+is a convenience view; `record_links.tsv.gz` is the smaller association table. No new structure
+prediction or resampling was needed for this migration.
+
+For strict paired-junction/peptide matching, use `matches_original` or `matches_repaired`.
+For agreement on that pair plus recorded species and both MHC chains/class, require
+`exact_tcr_pmhc_match`. `link_basis` distinguishes historical repair and accession provenance
+from literal sequence matches. Association does not establish identical full receptor constructs.
+The `cdr3` fields use VDJdb's anchor-inclusive junction convention, not anchor-excluding CDR3.
+Withdrawn assignments have `experimental_negative=false`; their receptors can have valid
+observations against other peptides. The independent audit found no exact paired-junction/peptide
+match to any current positive record for these 2,811 structures, across both original and repaired
+states. Receptors of 174 hashes occur against other peptides, as recorded in the audit files.

data/pdb_sources/README.md

@@ -5,3 +5,6 @@ These 32 gzip-compressed mmCIF files are complete, unmodified PDB downloads: the
 Coordinates remain in the deposited frame with original chain IDs, complete polymer sequences and chemical-residue names. These files have not been converted into the historical canonical orientation or assigned model confidence/contact metrics. `polymer_sequences.tsv.gz` preserves complete deposited chemistry independently of coordinate omissions; the PDB canonical sequence is a separate field and must not substitute for the literal chemical sequence.
 
 `SOURCES.json` records successful HTTP responses and hashes for all 32 coordinate files and all 32 entry-metadata responses. An input can be fetched again from its recorded `source_url` with `curl --fail --location URL --output FILE`; the SHA256 must match the manifest before treating the snapshot as identical.
+
+
+See [current record links](../../metadata/current/README.md) and the [database build-output contract](https://docs.isalgo.dev/vdjdb-db/outputs.html#structure-files-and-current-record-cross-links) for joining these sources to current VDJdb IDs.

low_confidence/withdrawn_10x/README.md

@@ -15,3 +15,6 @@ Of the 2811 structures in the low-confidence withdrawn-10X collection, 2538 (90.
 
 
 Model confidence, contact and angle metrics, and peptide provenance are copied unchanged from the original metadata. The low-confidence label describes the withdrawn assignment, not structural model quality.
+
+
+See [current record links](../../metadata/current/README.md) and the [database build-output contract](https://docs.isalgo.dev/vdjdb-db/outputs.html#structure-files-and-current-record-cross-links) for joining these sources to current VDJdb IDs.

metadata/current/README.md

@@ -1,6 +1,6 @@
 # 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. This is a local migration; no upload or new database release was performed.
+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
 
@@ -28,7 +28,7 @@ The 13 unlinked native accessions are `6v0y`, `6v13`, `6v15`, `6v18`, `6v19`, `6
 - `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.
-- `SOURCES_local.json`: input origins, historical evidence and regeneration/verification commands.
+Regeneration and verification tools are in [`tools/`](../../tools/). Local execution provenance is retained outside published dataset files.
 
 ## Sequence states and association strength
 
@@ -64,3 +64,22 @@ No fuzzy sequence matching, peptide trimming, donor collapse, source repair, str
 [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.

metadata/native_review/README.md

@@ -19,3 +19,6 @@ The primary sources are [PMID33863750](https://pubmed.ncbi.nlm.nih.gov/33863750/
 The old model label for `9GV6` is HLA-A*02:792N, a co-matching sequence-derived assignment. It is retained in the historical metadata for traceability, and its disagreement with the current curated restriction remains explicit. A top groove-sequence hit does not establish an experimental allele. Likewise, the old `7BYD` human/HLA labels are inconsistent with the published macaque/Mamu lipopeptide complex and are not admitted into the current database.
 
 All 32 requested coordinate sources (these 14 and the 18 missing archive accessions) were parsed as one batch; TCR annotation used one call per organism and MHC annotation one batch call. Complete polymer sequences were checked independently of unresolved coordinate residues. No new binding experiment, model, mutation, contact metric or allele inference was generated to force an excluded complex into the positive database.
+
+
+See [current record links](../current/README.md) and the [database build-output contract](https://docs.isalgo.dev/vdjdb-db/outputs.html#structure-files-and-current-record-cross-links) for joining these sources to current VDJdb IDs.