Scholarlink

<p>Flavonoids, a ubiquitous class of plant polyphenolic compounds, are known for their wide spectrum of biological functions, exhibiting diverse physiological functions and possessing significant application value in pharmaceuticals, foods, and nutraceuticals. Thus, it is of great significance to conduct the toxicity assessment. However, it is impossible to perform the experimental testing for a vast number of flavonoid chemcials. In this case, in silico methods are promising to address this problem. In strict accordance with OECD principles, this study established quantitative structure–toxicity relationship (QSTR) models for predicting flavonoid acute intraperitoneal toxicity in mice by employing GA-MLR methodology. Read-Across (RA) methodology was employed to estimate the toxicity based on structural similarity. RASTR descriptors were then calculated and pooled together with QSTR descriptors to establish a q-RASTR model. Importantly, intelligent consensus modelling was implemented as another method to enhance model's stability and predictive performance. Finally, the optimal QSTR model satisfied rigorous internal and external validation benchmarks, with <i>R</i><sup>2</sup> = 0.7887, <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({Q}_{LOO}^{2}\)</EquationSource> </InlineEquation> = 0.7327, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({Q}_{Fn}^{2}\)</EquationSource> </InlineEquation> = 0.8521–0.8772, <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\({CCC}_{test}\)</EquationSource> </InlineEquation> = 0.9299. Based on three computational toxicology methods (QSTR, RA, and consensus modeling), the optimal model was consensus model 0 (average predictions). This model was then applied to predict the toxicity of a real external dataset lacking toxicity values. A comparative analysis with the predictions from an open-source VEGA tool was conducted to verify the applicability and predictive reliability of our model. This work offers mechanistic insights into the toxicological behavior of flavonoids and provides a rapid toxicity prediction tool for evaluating the safety of flavonoid-based chemicals.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From structural features to mouse acute intraperitoneal toxicity prediction: a triple computational toxicology approach for safety assessment of flavonoids

  • 期刊论文

  • 发表日期：

    2026年7月23日

查看全文
  • Yichen Yang,
  • Na Zhang,
  • Ting Ren,
  • Lijiao Zhao,
  • Rugang Zhong,
  • Ning Lin,
  • Guohui Sun

摘要

Flavonoids, a ubiquitous class of plant polyphenolic compounds, are known for their wide spectrum of biological functions, exhibiting diverse physiological functions and possessing significant application value in pharmaceuticals, foods, and nutraceuticals. Thus, it is of great significance to conduct the toxicity assessment. However, it is impossible to perform the experimental testing for a vast number of flavonoid chemcials. In this case, in silico methods are promising to address this problem. In strict accordance with OECD principles, this study established quantitative structure–toxicity relationship (QSTR) models for predicting flavonoid acute intraperitoneal toxicity in mice by employing GA-MLR methodology. Read-Across (RA) methodology was employed to estimate the toxicity based on structural similarity. RASTR descriptors were then calculated and pooled together with QSTR descriptors to establish a q-RASTR model. Importantly, intelligent consensus modelling was implemented as another method to enhance model's stability and predictive performance. Finally, the optimal QSTR model satisfied rigorous internal and external validation benchmarks, with R2 = 0.7887, \({Q}_{LOO}^{2}\) = 0.7327, \({Q}_{Fn}^{2}\) = 0.8521–0.8772, \({CCC}_{test}\) = 0.9299. Based on three computational toxicology methods (QSTR, RA, and consensus modeling), the optimal model was consensus model 0 (average predictions). This model was then applied to predict the toxicity of a real external dataset lacking toxicity values. A comparative analysis with the predictions from an open-source VEGA tool was conducted to verify the applicability and predictive reliability of our model. This work offers mechanistic insights into the toxicological behavior of flavonoids and provides a rapid toxicity prediction tool for evaluating the safety of flavonoid-based chemicals.

Graphical abstract