Derived lineage and susceptibility tables for amr-clonalshare 1.0.0: cross-species atlas, veterinary two-resolution atlas, estimator benchmark and comparator arms

Abstract

Every table a number in the manuscript "amr-clonalshare: measuring how much of antimicrobial resistance travels with the clone" was read back from, at the version it was read at. Four groups of derived tables, plus the material needed to check them. Cross-species atlas. Species by agent cells built from NCBI Pathogen Detection isolate metadata, susceptibility calls and SNP cluster assignments, each with the out-of-sample clonal share, the realised share, both intervals, the estimability flag and a permuted-label control. Veterinary two-resolution atlas. The same source cut by host species and by sampling matrix, every cell run at two definitions of a lineage, the SNP cluster and the serovar, with the permuted-label control at both, plus the host and matrix classification rules and the per-cohort serovar composition. Estimator benchmark. Bias, root mean squared error, interval width and empirical coverage for five estimators scored against two estimands over a grid of lineage counts, cluster sizes, imbalance and true share. Comparator arms. Population-structure-only classification accuracy with its majority-class baseline, and the anchor verifying the restricted maximum likelihood arm against an independent derivation from the model statement. The archive also carries the pre-registrations written before each validation campaign was run, the analysis scripts that produced the tables, the validation ledger that records the digest of every evidence file, the supplementary material of the article, a README describing the layout, and a SHA-256 manifest of every file. Under evidence/E/ it carries the campaign trees that the validation ledger addresses by the E/ prefix, as the runs wrote them: the run receipts, the per-campaign results, the registration hashes and the test-suite records. The isolate-level source is public. What is deposited here is the derived join and the computed tables, because those are what a reader has to hold to check a number rather than recompute the retrieval. The software that produced them is archived at https://doi.org/10.5281/zenodo.22306354 and developed at https://github.com/maciejkochanowski/amr-clonalshare.