# Asian HLA benchmark data

## 1. Get the benchmark truth and its supporting data

**[Reference release 2026-10-20](https://bio2vec.net/data/Asian-HLA-Consortium/releases/2026-10-20/README.md)**

- [Truth-label TSV](https://bio2vec.net/data/Asian-HLA-Consortium/releases/2026-10-20/truth/haplotype_assignments.tsv) and [evidence/column guide](https://bio2vec.net/data/Asian-HLA-Consortium/releases/2026-10-20/truth/README.md).
- [Extracted regions and source annotations](https://bio2vec.net/data/Asian-HLA-Consortium/releases/2026-10-20/inputs/README.md).
- [Download the complete benchmark](https://bio2vec.net/data/Asian-HLA-Consortium/downloads/asian-hla-benchmark-2026-10-20.tar.gz), then [verify its checksum](https://bio2vec.net/data/Asian-HLA-Consortium/downloads/SHA256SUMS).

147 donors; eight HLA loci; 2,351 available gene copies. These are assembly-derived reference-label candidates for the benchmark under construction. Independent raw-read/phase certification is not yet complete. The release preserves the supported evidence level and ambiguity; caller predictions do not define truth.

## 2. Reproduce or evaluate

[Code and reproduction instructions](https://github.com/Asian-HLA-Consortium/hla-assembly-benchmark) (private repository).

[Method predictions and scores](https://bio2vec.net/data/Asian-HLA-Consortium/predictions/README.md) are separate immutable snapshots. Use the truth release, declared endpoint resolution, eligibility and donor coverage when comparing methods. An unfinished run is not an incorrect call.

## Layout

```
releases/2026-10-20/
  truth/          reference labels and sequence-matching evidence
  inputs/         extracted sequences, source annotations and mappings
  metadata/       sample sources, reference pins, expected result hashes
  provenance/     public source-binding evidence
  SHA256SUMS      checksums for this release

downloads/        complete benchmark archives and checksums
predictions/      caller predictions, scores and per-snapshot provenance
experiments/      completed graph/reference experiments and execution evidence
```

There are no moving aliases in this navigation. Cite a specific release/snapshot and its checksum. Historical download URLs remain available for existing scripts but are not the recommended starting point.

[Google benchmark tracking Sheet](https://docs.google.com/spreadsheets/d/18F0YMcjL8NUnPrjghI8z567m0oV4vYD7vHiM6KFASGo/edit)

## Results and reproducibility

[Results and interpretation](RESULTS.md) | [Software and database versions, source pins and checksums](SOFTWARE_DATABASES.json) | [Experiment release 2026-10-21](experiments/2026-10-21/).

OptiType, SpecHLA and HLA*LA are complete for all 86 evaluation donors. The Google Sheet includes Results_20261021, Baselines_20261021, Experiments_20261021, Execution_audit_20261021 and Versions_20261021 tabs.

## Simulated cohort extension

[Simulation protocol and reproducible profiles](SIMULATION.md): 52 Arab and 9 Saudi donors, NovaSeq X Plus–calibrated PE150, nominal 30× total coverage (15× per haplotype). Assembly preparation is running; simulation and baseline results are pending. This cohort is separate from the completed real-read benchmark.