Pan-disease cross-cohort comorbidity benchmark: degree-preserving null and LOPO oracle
Description
This dataset contains original analysis code, derived summary results, quality-control records, and source metadata for a secondary analysis of publicly available aggregate resources. The study benchmarks cross-cohort comorbidity transfer using FinnGen R14 and FinRegistry summaries from Risteys, an Austrian clinical comorbidity network, a runtime-enforced Austrian degree oracle, and a leave-one-pair-out diagnostic. The package includes the Finnish-to-Austrian benchmark, original and LOPO Austrian degree-oracle results, reproducibility scripts, aggregate audit results, a data dictionary, source provenance, and quality-control records. No individual-level participant records are included or were accessed.
Files
Steps to reproduce
The complete benchmark requires retrieval of the source resources under their original terms. Source resources: Risteys v2.8.0 / FinnGen R14 (https://risteys.finngen.fi/); EFO v3.92.0 (https://www.ebi.ac.uk/efo/); Mondo release 2026-07-06 (https://mondo.monarchinitiative.org/); Austrian clinical network, Figshare version 2 (https://doi.org/10.6084/m9.figshare.27102553.v2). Run the Python scripts from the code directory after rebuilding the excluded input tables. The LOPO diagnostic uses 5,000 degree-preserving permutations with seed 20260720.