Scholarlink

<p>Chemical graph theory offers a mathematical foundation for representing molecular structures and predicting their physicochemical properties. In this study, we employ generalized sum connectivity indices derived from hydrogen-suppressed molecular graphs of asthma drug molecules to perform quantitative structure–property relationship (QSPR) analysis. Topological descriptors are computed and used to model boiling point, critical temperature, critical volume, CLogP, and LogP. Baseline linear and quadratic regression models are established, after which machine learning techniques–Random Forest, XGBoost, and artificial neural networks—are applied to enhance predictive accuracy. Model robustness is evaluated through five-fold cross-validation and bootstrapping, with performance assessed using standard error metrics. Comparative results demonstrate that XGBoost consistently achieves the highest predictive accuracy, yielding, for example, <InlineEquation ID="IEq1"><EquationSource Format="TEX">\(r^2 = 0.9967\)</EquationSource></InlineEquation> for critical volume and <InlineEquation ID="IEq2"><EquationSource Format="TEX">\(r^2 = 0.9879\)</EquationSource></InlineEquation> for boiling point. The integration of generalized sum connectivity indices with advanced machine learning provides a powerful and interpretable framework for molecular property prediction, underscoring the value of chemical graph theory in computational drug design.</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

Exploring physicochemical property interplay in asthma drug molecules via connectivity-based graph descriptors and learning models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月22日

查看全文
  • Sadia Noureen,
  • Nazma Ashraf,
  • Saood Azam,
  • Eman Fatima,
  • Adnan Aslam,
  • Keneni Abera Tola

摘要

Chemical graph theory offers a mathematical foundation for representing molecular structures and predicting their physicochemical properties. In this study, we employ generalized sum connectivity indices derived from hydrogen-suppressed molecular graphs of asthma drug molecules to perform quantitative structure–property relationship (QSPR) analysis. Topological descriptors are computed and used to model boiling point, critical temperature, critical volume, CLogP, and LogP. Baseline linear and quadratic regression models are established, after which machine learning techniques–Random Forest, XGBoost, and artificial neural networks—are applied to enhance predictive accuracy. Model robustness is evaluated through five-fold cross-validation and bootstrapping, with performance assessed using standard error metrics. Comparative results demonstrate that XGBoost consistently achieves the highest predictive accuracy, yielding, for example, \(r^2 = 0.9967\) for critical volume and \(r^2 = 0.9879\) for boiling point. The integration of generalized sum connectivity indices with advanced machine learning provides a powerful and interpretable framework for molecular property prediction, underscoring the value of chemical graph theory in computational drug design.