> Website excerpt of FORMAT_COVERAGE_20260915.md, through the full-file evidence section. Historical aggregate framing removed under the 2026-09-25 claim audit.

# Full-file format-coverage register (updated 2026-09-23)

## Scope

This register separates source acquisition from harmonization evidence. The local
100k/10k files are preliminary format gates only. They are not full-file paper
results and must not be counted as independent studies. Full harmonization is
allowed only after the corresponding format gate, scientific metadata
confirmation, and full-file terminal evidence pass.

## Authoritative full sources acquired

| Source format | Dataset and independence unit | Exact local rows | Chromosome/contig labels | Local source | Published source | Integrity evidence | Current disposition |
|---|---|---:|---:|---|---|---|---|
| Pan-UKBB per-phenotype flat file | Standing height, cross-population high-quality meta-analysis (`continuous-50-both_sexes-irnt`) | 28,987,534 | 23 | `/Users/mlemsalu/Desktop/gwas-harmonizer/data/full_input/PanUKBB_continuous-50-both_sexes-irnt.full.tsv.bgz` | [Pan-UKBB downloads](https://pan.ukbb.broadinstitute.org/downloads/index.html) and phenotype manifest | MD5 `d883d1ec0840c3fec63457648ef0cb2f` (matches manifest); SHA-256 `4e5d5be06d2dff77e842751b3bca37bcb2ecca2603702f92ce101b1ee45d51ac`; gzip test passed | Both run `d1177b0f-1015-4db8-97b1-2a0f895a3e03` and replacement `360d3d79-9776-4ddc-af51-a325d9e9417c` lost download status and produced no local evidence bundle. Production logs show the latter completed a 28,968,245-row export before cleanup deleted its ready job at upload-session expiry. Backend fix `b18342d` is deployed, but a 2026-09-23 connector start reused the old run ID and failed with `Original uploaded file not found`; no fresh run was created. Provisional counts remain excluded. |
| GCTA fastGWA | UK Biobank standing height trait 50 | 8,531,093 | 22 | `/Users/mlemsalu/Desktop/gwas-harmonizer/data/full_input/fastGWA_UKBB_standing_height_50.v1.1.full.fastGWA.gz` | [GCTA data documentation](https://yanglab.westlake.edu.cn/software/gcta/static/gcta_doc_latest.pdf) and `UKB_impute_v1.1.csv` | MD5 `4bd51bcc975c001d2e5958526d459c51`; SHA-256 `3e6121eb2bb45b6f0591fa143692f08a8566fa7532dfa5c2332286c7f359330c`; gzip test passed | Corrected full run `d5fb897e-e0e3-4b8c-b75e-97f9c9811280` completed: 7,861,307 retained and 669,786 quarantined. Its local evidence bundle records the exact blockers and two source p-value-floor inconsistencies; it is not submission-ready. |
| PLINK 2 linear regression | UK Biobank standing height benchmark, prespecified training fold 1; folds are not independent studies | 1,074,448 | 22 | `/Users/mlemsalu/Desktop/gwas-harmonizer/data/full_input/PLINK2_UKBB_standing_height_fold1.full.glm.linear.gz` | [Zenodo record 14270953](https://zenodo.org/records/14270953), archive `benchmark_sumstats.tar.gz` | Archive MD5 `dc730082cbc31d3f1fa41c64ab6c6b9d` (matches Zenodo); archive SHA-256 `be550c69a6150bee8343c957ed440953e0690b7e68f4246623955d004afdc1c9`; combined fold SHA-256 `19d38fa30bf763be271c189f8bb847211eddb737219c1b1ff4e2e7f4d1c86477`; gzip test passed | Full run `dd52a38e-48c6-412f-8893-7428c7a03863` completed: 1,073,599 retained and 849 quarantined. The harmonization audit is complete with no severe p-value inconsistencies, but the canonical bundle is not a GWAS Catalog submission package. |

The fastGWA exact file count is reported separately from the resource-level count
in the GCTA documentation. No attempt is made to force those two counts to agree.

## Independently verified statistical semantics

The three full runs used the same explicit declarations: quantitative trait,
additive beta, and standard error on the beta scale. These declarations were
verified from source documentation rather than inferred from filenames:

- Pan-UKBB documents `beta_meta_hq` as the estimated alternate-allele effect and
  `se_meta_hq` as its standard error for the quantitative `continuous-50`
  phenotype ([per-phenotype file documentation](https://pan.ukbb.broadinstitute.org/docs/per-phenotype-files/index.html)).
- GCTA identifies phenotype 50 as standing height and defines fastGWA `A1` as
  the effect allele, `BETA` as the SNP effect and `SE` as its standard error
  ([GCTA documentation](https://yanglab.westlake.edu.cn/software/gcta/static/gcta_doc_latest.pdf)).
- PLINK 2 `.glm.linear` is linear regression for a quantitative phenotype; `A1`
  is the tested/effect-dosage allele, `BETA` is the regression coefficient and
  `SE` is its standard error
  ([PLINK 2 association documentation](https://www.cog-genomics.org/plink/2.0/assoc)).

This confirms scale compatibility; it does not by itself establish cohort,
phenotype, ancestry or association-model comparability across the three files.

## Full-file outcomes (verified status 2026-09-23)

| Dataset | Input | Retained | Quarantined | Other removed | Retention | Liftover failures | Severe p-value inconsistencies | Full-run interpretation |
|---|---:|---:|---:|---:|---:|---:|---:|---|
| Pan-UKBB | 28,987,534 | — | — | — | — | — | — | No locally verifiable full-file bundle exists. The backend lease/retention bug is fixed, but the connector has not created a fresh run; preview and production-log counts are excluded from paper results. |
| fastGWA | 8,531,093 | 7,861,307 | 669,786 | 0 | 92.1489% | 7,232 | 2 | Corrected full-file evidence supersedes the stale 6,662,386-row result. The two severe rows are source p-value floor cases near `1e-308`, not evidence that beta or SE use a different scale. |
| PLINK2 fold 1 | 1,074,448 | 1,073,599 | 849 | 0 | 99.9210% | 562 | 0 | Full harmonization audit completed; 849 rows were quarantined across unmapped, non-primary, final-reference, ambiguous-reference and orientation-invariant categories. |

Exact row accounting holds for the completed, locally evidenced runs. The new files extend format coverage;
they are not counted as additional independent studies. The legacy aggregate benchmark was withdrawn; the current claim audit governs public interpretation.

Local evidence for the fastGWA and PLINK2 runs:

- `/Users/mlemsalu/Desktop/gwas-harmonizer/results/benchmarks/format_coverage_20260915/full/fastGWA/fastGWA_UKBB_standing_height_full-file_format_benchmark_gwasharmonizer_20260917-093502/` — corrected package SHA-256 `9fbbe2a3fc9bcf5361e9f585d8e512384307d6a9e1f67a2795ad0b324ffd9f61`; manifest SHA-256 `1ffd3af2e723d5f1df9fd500315d13d847f0eeed95a2963e99f7ca3986a91756`.
- `/Users/mlemsalu/Desktop/gwas-harmonizer/results/benchmarks/format_coverage_20260915/full/PLINK2_fold1/PLINK2_UKBB_standing_height_fold_1_full-file_format_benchmark_gwasharmonizer_20260917-102122/` — package SHA-256 `f69ceb0ee34347c72cc1de62767819bf9a11c303253c8d313913b23a9b8dde5a`; manifest SHA-256 `839713f2f3183369e1a962f886e60fd5bb0e3247ecedff32efdc56c66de747af`.