# Supplementary Table S1. Validation datasets and reproducibility register

## Scope and use

This is the submission working register for the four public subset-scale runs
described in the working manuscript. It separates verified local evidence from
metadata that still requires an author/source check. It does **not** represent
the withdrawn 54.9M aggregate, the incomplete Pan-UKBB comparison, or a claim
of full-file accuracy.

Each included result must remain traceable to its local `evidence_manifest.json`
and `paper_metrics.csv`. `SUBMISSION_READY=false` is a product readiness
assessment under the selected profile; it is not an association-validity or
biological-validity verdict.

| Dataset / role | Exact input and SHA-256 | Declared semantics and build | Terminal evidence | Input → retained | Paper use | Outstanding metadata / limitation |
|---|---|---|---|---:|---|---|
| FinnGen R13 T2D, clean biobank subset | `summary_stats_finngen_R13_T2D_100k.gz`<br>`abf3a414ed2abbd1b6eaf1ec6fd6efbfe492ca5049fb7fe81dff4ca3a15ea9b5` | Binary; log-odds effect and same-scale SE; GRCh38 | [`evidence manifest`](../../results/benchmarks/GWAS%20100k%20row%20files/result/FinnGen_R13_T2D_100k_gwasharmonizer_20260729-143036/evidence_manifest.json) · [`metrics`](../../results/benchmarks/GWAS%20100k%20row%20files/result/FinnGen_R13_T2D_100k_gwasharmonizer_20260729-143036/paper_metrics.csv) | 99,999 → 99,999 | Scoped configured-behavior and audit example | [FinnGen download documentation](https://github.com/FINNGEN/finngen-documentation/blob/master/data-download.md) was checked 2026-09-26 and confirms GRCh38, ALT-effect and beta/SE conventions. The exact R13 T2D source object, byte size, access date and redistribution terms remain unavailable from the local subset and must be recovered through FinnGen’s access route. Original package records unresolved conflicting rsIDs and incomplete audit trail; do not call submission-ready. |
| Neale Lab UKB height 50_irnt, liftover subset | `Neale Lab UK Biobank standing height 100k.tsv.bgz`<br>`69e8bde0cb41642fbd05da3d6094e1605d5e3d102d03d1b20c143621db4b8f9b` | Quantitative; additive beta and same-scale SE; GRCh37 → GRCh38 | [`evidence manifest`](../../results/benchmarks/GWAS%20100k%20row%20files/Neale_swap_resolution_rerun_20260817/Neale_Lab_UK_Biobank_standing_height_100k_swap_resolution_verified_rerun_gwasharmonizer_20260817-122507/evidence_manifest.json) · [`metrics`](../../results/benchmarks/GWAS%20100k%20row%20files/Neale_swap_resolution_rerun_20260817/Neale_Lab_UK_Biobank_standing_height_100k_swap_resolution_verified_rerun_gwasharmonizer_20260817-122507/paper_metrics.csv) | 100,000 → 99,829 | Scoped liftover/audit example | [Official object](https://broad-ukb-sumstats-us-east-1.s3.amazonaws.com/round2/additive-tsvs/50_irnt.gwas.imputed_v3.both_sexes.tsv.bgz) checked 2026-09-26: 539,684,138 bytes; S3 ETag `c46ce2d16c420f61c9f3d656167f882c-65` (multipart ETag, not an upstream file checksum). Still blocked for the selected profile by ambiguous CI handling and missing GWAS Catalog schema request. |
| Suzuki T2DGGI EAS, ancestry-aware liftover subset | `Suzuki.Nature2024.T2DGGI.EAS.sumstats_100k.zip`<br>`baeb0c78194476c0fed43f1f0f1a0ec48e5af4c6944838a28a03ed25b5644f4c` | Binary; log-odds effect and same-scale SE; EAS; GRCh37 → GRCh38 | [`evidence manifest`](../../results/benchmarks/GWAS%20100k%20row%20files/Suzuki_swap_resolution_rerun_20260817/Suzuki_Nature_2024_T2DGGI_EAS_100k_swap_resolution_verified_rerun_gwasharmonizer_20260817-122852/evidence_manifest.json) · [`metrics`](../../results/benchmarks/GWAS%20100k%20row%20files/Suzuki_swap_resolution_rerun_20260817/Suzuki_Nature_2024_T2DGGI_EAS_100k_swap_resolution_verified_rerun_gwasharmonizer_20260817-122852/paper_metrics.csv) | 100,000 → 92,022 | Scoped EAS-panel, liftover and conservative-quarantine example | [Official DIAGRAM/T2DGGI landing page](https://www.diagram-consortium.org/downloads.html) checked 2026-09-26; it publishes SHA1 `9f65d8f3c20ac0a39d004f6ab7c3ffdfd760b914` and MD5 `4f18dd6bcc9b7dfd0c54f2dea3bf2a09` for the EAS archive. The form-mediated resolved object URL, byte size and original access timestamp remain to be captured. Two severe p-value inconsistencies and ambiguous CI handling remain explicit blockers. |
| GCST90018642 hypothyroidism, repository-harmonized control | `GWAS_Catalog_100k_GCST90018642_harmonised_control.tsv.gz`<br>`a3a6f876f9395d3aaff111799669bae4c0232c64d61b145d9b5d651d52df585d` | Binary; log-odds effect and same-scale SE; GRCh38 | [`evidence manifest`](../../results/benchmarks/GWAS%20100k%20row%20files/result/GCST90018642_harmonised_control_100k_final_four-file_test_gwasharmonizer_20260729-144724/evidence_manifest.json) · [`metrics`](../../results/benchmarks/GWAS%20100k%20row%20files/result/GCST90018642_harmonised_control_100k_final_four-file_test_gwasharmonizer_20260729-144724/paper_metrics.csv) | 100,000 → 83,382 | Scoped conservative-policy control | [Current official harmonized object](https://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90018001-GCST90019000/GCST90018642/harmonised/GCST90018642.h.tsv.gz) and [metadata](https://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90018001-GCST90019000/GCST90018642/harmonised/GCST90018642.h.tsv.gz-meta.yaml) checked 2026-09-26: 561,748,458 bytes, MD5 `73e4e5dc395c3c2135697f87bda13fa5`, GRCh38, East Asian sample ancestry. This matches the recorded local full-source MD5 and SHA-256 `7f0e747b719a54dbfc81ba07ecbe7568962687b7ff8be06def06f79f266bc2e4`. Unsupported palindromes and ambiguous indels were quarantined. |

## Reference resources and settings

All four runs used `target_build=GRCh38`, reference alignment, missing-rsID
handling and conservative palindromic policy. The manifest records the precise
options and package artifacts; the final Zenodo archive must also include the
FASTA, chain and rsID resource version/checksum manifest. Do not claim that
identical options imply comparability of study populations or association
models.

## Release checklist

- Record each upstream object URL, access date, byte size and checksum from the
  source release, not merely its local subset.
- Copy the referenced terminal evidence packages and this table into the
  release archive, retaining their SHA-256 values.
- Update this table only after the authoritative Word manuscript and live
  draft agree on the selected run IDs and scope.
