2026-10-11T18:20:32Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33794058 2026-09-15T12:55:02Z category_146 category_7 category_19 category_873 category_734 category_64 portal_87 item_type_3 month_year_09_2026

A dose-unit-aware predictive toxicology framework for literature-derived in vitro carbon nanotube cytotoxicity Deniz Özkan Vardar (24934119) Necati Vardar (24934122) Space Science Medicine Pharmacology Chemical Sciences not elsewhere classified Biological Sciences not elsewhere classified Cancer Carbon nanotubes predictive toxicology computational toxicology nanotoxicology in vitro cytotoxicity Carbon nanotubes (CNTs) are increasingly used as engineered nanomaterials, but their in vitro cytotoxicity remains difficult to synthesize because reported viability outcomes vary across dose metrics, exposure durations, CNT types, functionalization states, cell models, and assay systems. This study developed an exploratory, literature-derived, dose-unit-aware predictive toxicology framework for binary classification of CNT-induced cytotoxicity using standardized cell viability data. Experimental-condition-level records were extracted from eligible in vitro CNT studies and harmonized according to dose, dose unit, exposure duration, CNT type, functionalization, cell line, assay type, and viability outcome. Cytotoxicity was defined as cell viability below 70%. A primary-core dataset, including 146 µg/mL records from 11 articles, was used for nominal concentration-based modeling, whereas a secondary unit-aware dataset, including 212 records from 13 articles, incorporated both µg/mL and µg/cm2 observations while retaining dose-unit identity as an explicit predictor. Logistic Regression and Random Forest classifiers were trained using log-transformed dose-related variables and experimental-context predictors. Performance was assessed using stratified cross-validation, article-grouped cross-validation, leave-one-article-out validation, and leave-one-family-out validation with fold-level variability considered to reflect uncertainty and to mitigate overoptimistic estimates arising from within-study similarity. Random Forest showed higher point estimates under conventional stratified validation, whereas Logistic Regression provided comparatively stable discrimination under stricter grouped-validation settings. Ablation-based stress testing identified high-dose nontoxic A549/MTT observations as a major source of false-positive toxicity predictions, particularly for Random Forest. A dose × dose-unit interaction sensitivity analysis did not consistently improve grouped or leave-one-out validation; therefore, the additive unit-aware model was retained as the primary unit-aware analysis. The unit-aware analysis demonstrated that surface-dose records can be incorporated without assuming equivalence between nominal mass concentration and surface-area-normalized exposure. The framework provides a transparent strategy for curating, validating, stress-testing, and cautiously modeling heterogeneous literature-derived CNT viability data in nanosafety research. 2026-09-15T12:55:02Z Dataset Dataset 10.6084/m9.figshare.33794058.v1 https://figshare.com/articles/dataset/A_dose-unit-aware_predictive_toxicology_framework_for_literature-derived_i_in_vitro_i_carbon_nanotube_cytotoxicity/33794058 CC BY 4.0