Ting Ren

With the increasing environmental presence of pharmaceuticals and industrial chemicals, rapid and reliable tools are urgently needed to assess potential risks to ecosystems and human health. However, experimental acute oral toxicity data for aniline compounds and their derivatives (AnCDs) remain scarce, limiting large-scale hazard evaluation. Quantitative structure–toxicity relationship (QSTR) modeling offers an efficient alternative to experimental testing and supports regulatory risk assessment. In this study, acute oral toxicity data (LD50) for 54 AnCDs in rat and 57 AnCDs in mouse were collected to develop predictive models in compliance with OECD guidelines. The resulting QSTR models demonstrated robust internal validation (R2 > 0.6 and [[EQUATION]] > 0.6) and strong external predictive performance ([[EQUATION]] > 0.7, [[EQUATION]] > 0.7, and CCCtest > 0.85). To further enhance predictive capability, a novel quantitative read-across structure–toxicity relationship (q-RASTR) approach integrating read-across information with 2D molecular descriptors was developed, yielding enhanced external predictivity. Mechanistic interpretation revealed that valence electron distribution, topological distances between atomic pairs, and key electronic and structural molecular features play critical roles in determining acute oral toxicity. To our knowledge, this is the first study to integrate QSTR and q-RASTR as new approach methodologies (NAMs) for predicting rodent acute oral toxicity of AnCDs. Using atomic-centered fragments (ACFs), the validated models were applied to hundreds of untested external compounds, enabling toxicity prioritization and the identification of high-risk substances. This dual-model NAMs framework provides a valuable tool for chemical risk assessment, regulatory decision-making, and the rational design of safer chemicals and pharmaceuticals.