Enhancing interpretability in toxicity prediction: a reliable multilinear regression model for hydrazine derivatives

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The accurate assessment of rat oral toxicity for hydrazine derivatives is essential for ensuring chemical safety and regulatory compliance. To this end, a multiple linear regression model was developed to predict log L D 50 values using an experimental dataset of 77 compounds (comprising 58 training, 14 test, and 5 external validation samples), with internal reliability substantiated by leave-one-out cross-validation. The model demonstrates high predictive performance, evidenced by a determination coefficient ( R 2 ) of 0.8202 and a strong linear correlation between predicted and experimental values. Compared to more complex computational approaches, this multilinear regression framework offers distinct advantages in model transparency, structural interpretability, and computational efficiency. By minimizing reliance on opaque fitting procedures, the model effectively mitigates the risk of overfitting while providing reliable and actionable toxicity estimates within its defined applicability domain. These results validate the proposed methodology as a precise and efficient tool for the reliable screening of hydrazine derivatives, supporting safer chemical development and risk assessment strategies.