Scholarlink

Background <p>Agrochemical residues and naturally occurring toxins in vegetables, particularly tomatoes, pose significant risks to human health and the environment. Traditional toxicity testing methods are resource-intensive and reliant on animal models. In response, <i>in-silico</i> approaches like QSAR (Quantitative Structure–Activity Relationship) and q-RASAR (Quantitative Read-Across Structure–Activity Relationship) have emerged as reliable alternatives for toxicity prediction and data gap filling.</p> Method <p>This study utilized EC<sub>25</sub> toxicity data for tomato-related agrochemicals, curated from the OPP database. After structural refinement and descriptor calculation using PaDEL, QSAR and q-RASAR models were developed using multiple linear regression (MLR) in QSARINS. Model performance was validated through OECD-compliant statistical metrics. An Intelligent Consensus Predictor (ICP) approach was employed to enhance prediction accuracy by integrating outputs from all models. Applicability Domain (AD) and Prediction Reliability Index (PRI) were assessed to ensure model robustness.</p> Result <p>QSAR Models 1 and 2 and q-RASAR Models 3 and 4 showed strong statistical performance with R<sup>2</sup> values &gt; 0.75, Q<sup>2</sup>LOO &gt; 0.63, Q<sup>2</sup><sub>F1</sub>- Q<sup>2</sup><sub>F2</sub> &gt; 0.557 and CCC<sub>ext</sub> = 0.718. Key molecular descriptors such as ATSC0c, ATSC5p, and VCH-5 were identified as toxicity-determining features. Consensus-based validation revealed QSAR Model M2 as the most predictive, with high reliability across both ECOTOX and PPDB (Footprint) databases. It successfully predicted toxicity for 612 and 1,781 unknown compounds, respectively, classifying them as toxic or non-toxic and flagging structurally critical alerts.</p> Conclusion <p>The developed models provide an efficient, animal-free framework for predicting agrochemical toxicity in tomatoes, aiding regulatory risk assessment and supporting the design of safer, sustainable agrochemicals.</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

QSAR, q-RASAR and intelligent consensus-based predictions of agrochemicals for Tomato (Solanum lycopersicum) phytotoxicity: risk assessment, food safety and sustainability

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Surbhi Jyoti,
  • Souranava Jana,
  • Anjali Murmu,
  • Balaji Wamanrao Matore,
  • Jagadish Singh,
  • Partha Pratim Roy

摘要

Background

Agrochemical residues and naturally occurring toxins in vegetables, particularly tomatoes, pose significant risks to human health and the environment. Traditional toxicity testing methods are resource-intensive and reliant on animal models. In response, in-silico approaches like QSAR (Quantitative Structure–Activity Relationship) and q-RASAR (Quantitative Read-Across Structure–Activity Relationship) have emerged as reliable alternatives for toxicity prediction and data gap filling.

Method

This study utilized EC25 toxicity data for tomato-related agrochemicals, curated from the OPP database. After structural refinement and descriptor calculation using PaDEL, QSAR and q-RASAR models were developed using multiple linear regression (MLR) in QSARINS. Model performance was validated through OECD-compliant statistical metrics. An Intelligent Consensus Predictor (ICP) approach was employed to enhance prediction accuracy by integrating outputs from all models. Applicability Domain (AD) and Prediction Reliability Index (PRI) were assessed to ensure model robustness.

Result

QSAR Models 1 and 2 and q-RASAR Models 3 and 4 showed strong statistical performance with R2 values > 0.75, Q2LOO > 0.63, Q2F1- Q2F2 > 0.557 and CCCext = 0.718. Key molecular descriptors such as ATSC0c, ATSC5p, and VCH-5 were identified as toxicity-determining features. Consensus-based validation revealed QSAR Model M2 as the most predictive, with high reliability across both ECOTOX and PPDB (Footprint) databases. It successfully predicted toxicity for 612 and 1,781 unknown compounds, respectively, classifying them as toxic or non-toxic and flagging structurally critical alerts.

Conclusion

The developed models provide an efficient, animal-free framework for predicting agrochemical toxicity in tomatoes, aiding regulatory risk assessment and supporting the design of safer, sustainable agrochemicals.