Dataset for: Structural Constraints on Material-Identity-Aware Validation in Literature-Derived Nanotoxicology Models

This dataset accompanies the manuscript “Structural Constraints on Material-Identity-Aware Validation in Literature-Derived Nanotoxicology Models”. The data package contains manuscript-facing processed datasets, audit outputs, frozen decision logs, material-identity summaries, publication-lineage diagnostics, feature-taxonomy summaries, article-grouped out-of-fold model-output summaries, paired article-cluster bootstrap summaries, calibration summaries, and top-risk operating-point diagnostics used to support the analyses reported in the manuscript. The study evaluates whether material-identity-aware validation can be structurally established before model interpretation in two literature-derived nanotoxicology corpora: an in vitro carbon nanotube cytotoxicity corpus and a polystyrene nanoparticle/microplastic cytotoxicity corpus. The package is intended to support transparency, reproducibility, and traceability of the frozen audit workflow, identity-feasibility decisions, leakage-control structure, and clustered model-evaluation summaries. The package includes a README file, data dictionary, manifest with SHA-256 checksums, manuscript-facing aggregate tables, supplementary information document, and source workbooks used to assemble the reported summaries. Some source-level materials derived from copyrighted publications are not redistributed; where applicable, only processed or manuscript-facing outputs are included.