Type 2 diabetes–conditioned NMR screening reveals a shared circulating metabolic program aligned with postprandial hepatic insulin resistance

This record contains the analysis code and reproducibility materials accompanying the study of a type 2 diabetes-conditioned circulating NMR genetic-effect pattern. The archive includes retained NMR–FinnGen variant inputs, numerical supplementary tables, reviewed aggregate HRS results, source-access information, and computational reconstruction receipts. The local package version is 2026-10-04-v6; the Zenodo record was published on 4 October 2026.

The reconstructions include outcome matching for selected MR inputs, all 3,289 cells of the 299-locus by 11-trait NMR matrix, a same-input A294 centre-fitting replay, and scoped GCKR A381/A382/A397 replays. 

The archive excludes manuscript and supplementary narrative, completed reporting checklists, finished publication artwork, HRS participant-level records, MVP variant-level coefficients, and reference participant genotypes. Readers must obtain externally held data under the relevant providers' access conditions. Author software is licensed under MIT; author-created documentation and aggregate results are licensed under CC BY 4.0. Third-party data retain their original terms, as documented in LICENSE_NOTICE.md and THIRD_PARTY_ATTRIBUTION.md.

Workbook annotation corrections identified on 4 October 2026

The following corrections apply to descriptive metadata in tables/Supplementary_Tables_S1-S7.xlsx within the archived ZIP:

1. README, row 11: the S7 workbook coverage is S7A–S7I, containing 322 data rows in total. The row should describe aggregate endpoints, state audit, out-of-bag calibration, synchronized gates, decision-curve analysis, participant flow, baseline characteristics, variable missingness and follow-up. The earlier range S7A–S7E and total of 23 refer only to the original five S7 sheets; S7F–S7I were already included in the archive.

2. S3D LocusGeometry299, cell A2: signed programme coordinates use the frozen allele convention and have no universal T2D-risk interpretation. C, A/N and R membership counts are pair-specific T2D-coupling annotations. Reference-allele sensitivity is reported in Table S9. The phrase "allele-neutral display coordinates" should not be applied to signed coordinates.

3. S3A Object, cell G5: the complete representative-variant selection order is descending selected-trait recurrence, descending summed exposure F statistic, descending mean absolute exposure Z statistic, and then ascending rsID. The earlier short description omitted the mean-absolute-Z tie breaker. The retained production script and Supplementary Methods S4.1 already specify the complete order.

These corrections change workbook navigation and explanatory labels only. They do not change analytical coefficients, variant selections, numerical result rows, or plotted results. The manuscript-submission workbook has been updated accordingly. The archive remains the preserved computational record; its pre-deposit wording about DOI assignment describes the preparation stage, whereas the public record now has the DOI given above.