Reproducibility Package: A Dependency-Aware Reliability Evaluation Procedure for Alignment-Free Classification of Giardia duodenalis Assemblages Across Studies and Biological Domains

This record contains the reproducibility and supplementary materials for the study “A Dependency-Aware Reliability Evaluation Procedure for Alignment-Free Classification of Giardia duodenalis Assemblages Across Studies and Biological Domains.” The study evaluates alignment-free classification of Giardia duodenalis assemblages using the beta-giardin (bg), glutamate dehydrogenase (gdh), and triosephosphate isomerase (tpi) loci. The final modeling dataset contains 9,261 records representing 7,411 locus-specific unique sequences. The evaluation procedure includes random record-level evaluation, exact-duplicate-disjoint evaluation, Strict-99 similarity-group-disjoint nested evaluation, Strict-98 sensitivity analysis, similarity-purged study and biological-domain holdouts, probability calibration, selective prediction, and an open-set challenge using assemblages G and H. The deposited package includes accession and source metadata, curated-data files, the audit of 449 records representing 256 unique sequences involved in direct cross-assemblage Strict-99 conflicts, Strict-99 and Strict-98 dependency assignments, fixed cross-validation and holdout definitions, analysis scripts, out-of-fold prediction files, held-out-study results, calibration and open-set summaries, supplementary tables, confusion matrices, and SHA-256 file manifests. Source nucleotide sequences are publicly available through GenBank. No new human or animal experimentation was performed. Version 1.0.0 corresponds to the reproducibility package prepared for the associated manuscript submission.