|国家预印本平台
首页|Rigorous Evaluation of Predictive Toxicity Models by Multi-Objective Optimization of Reference Compound Lists Using Genetic Algorithms

Rigorous Evaluation of Predictive Toxicity Models by Multi-Objective Optimization of Reference Compound Lists Using Genetic Algorithms

Rigorous Evaluation of Predictive Toxicity Models by Multi-Objective Optimization of Reference Compound Lists Using Genetic Algorithms

来源:Arxiv_logoArxiv
英文摘要

In pharmaceutical safety assessments, validation studies are essential for evaluating the predictive performance and reliability of alternative methods prior to regulatory acceptance. Typically, these studies utilize reference compound lists selected to balance multiple critical factors, including chemical structure, physicochemical properties, and toxicity profiles. However, the inherent trade-offs among these criteria complicate the independent optimization of each factor, necessitating a comprehensive multi-objective optimization approach. To address this challenge, we propose a novel multi-objective optimization framework employing a Genetic Algorithm (GA) to simultaneously maximize structural, physicochemical, and toxicity diversity of reference compound lists. Applying this methodology to existing validation study datasets, we demonstrated that GA-optimized compound lists achieved significantly higher overall diversity compared to randomly generated lists. Additionally, toxicity prediction models tested on GA-optimized compound lists exhibited notably lower predictive performance compared to random selections, confirming that these lists provide a rigorous and unbiased assessment environment. These findings emphasize the potential of our GA-based method to enhance the robustness and generalizability of toxicity prediction models. Overall, our approach provides valuable support for developing balanced and rigorous reference compound lists, potentially accelerating the adoption of alternative safety assessment methods by facilitating smoother regulatory validation processes.

Yohei Ohto、Tadahaya Mizuno、Yasuhiro Yoshikai、Hiromi Fujimoto、Hiroyuki Kusuhara

医学研究方法

Yohei Ohto,Tadahaya Mizuno,Yasuhiro Yoshikai,Hiromi Fujimoto,Hiroyuki Kusuhara.Rigorous Evaluation of Predictive Toxicity Models by Multi-Objective Optimization of Reference Compound Lists Using Genetic Algorithms[EB/OL].(2025-05-11)[2025-06-14].https://arxiv.org/abs/2505.07140.点此复制

评论