Scholarlink

Purpose <p>The transfer of pharmaceutical agents into breast milk poses a critical safety concern for breastfeeding infants, necessitating precise prediction of milk-to-plasma (M/P) concentration ratios for safe and effective pharmacotherapy during lactation. This study aims to develop an innovative Quantitative Structure-Property Relationship (QSPR) model to predict the logarithm of the M/P ratio (log(M/P)) using a curated dataset of 100 structurally diverse pharmaceuticals. The study explores whether QSPR can enhance early risk assessment, optimize lead compounds, and reduce reliance on resource-intensive experimental trials.</p> Methods <p>The QSPR model was constructed using Monte Carlo optimization, employing conformation-independent SMILES-based descriptors (both local and global) and graph-theoretical invariants, such as extended connectivity indices and path counts. The dataset was randomly divided into 75:25 training-test subsets across three independent runs, with validation conducted via internal cross-validation, external testing, y-randomization, and applicability domain analysis using the statistical defect approach.</p> Results <p>The models exhibited strong predictive performance, with training set coefficients of determination (R²) ranging from 0.8716 to 0.8808 and test set R² peaking at 0.9232. Additional metrics, including Concordance Correlation Coefficient (CCC up to 0.9366), Index of Ideality of Correlation (IIC up to 0.9608), and mean absolute error (MAE as low as 0.1657), confirmed accuracy and generalizability. SMILES-based analysis identified molecular fragments with positive and negative impact on log(M/P.</p> Conclusion <p>This study establishes conformation-independent QSPR modeling as an important tool for early-stage drug development and pharmacovigilance, providing actionable insights to minimize infant exposure and improve maternal and infant health outcomes for breastfeeding women through safer therapeutic options.&#xa0;</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

Prediction of Milk-to-Plasma Drug Concentration Ratios Using QSPR Modeling Based on Monte Carlo Optimization

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • Sonja Pop-Trajković Dinić,
  • Milan Trenkić,
  • Aleksandar Živadinović,
  • Predrag Vukomanović,
  • Milan Stefanović,
  • Dejan Mitić,
  • Jelena Milošević Stevanović,
  • Jelena Živković,
  • Aleksandar Veselinović

摘要

Purpose

The transfer of pharmaceutical agents into breast milk poses a critical safety concern for breastfeeding infants, necessitating precise prediction of milk-to-plasma (M/P) concentration ratios for safe and effective pharmacotherapy during lactation. This study aims to develop an innovative Quantitative Structure-Property Relationship (QSPR) model to predict the logarithm of the M/P ratio (log(M/P)) using a curated dataset of 100 structurally diverse pharmaceuticals. The study explores whether QSPR can enhance early risk assessment, optimize lead compounds, and reduce reliance on resource-intensive experimental trials.

Methods

The QSPR model was constructed using Monte Carlo optimization, employing conformation-independent SMILES-based descriptors (both local and global) and graph-theoretical invariants, such as extended connectivity indices and path counts. The dataset was randomly divided into 75:25 training-test subsets across three independent runs, with validation conducted via internal cross-validation, external testing, y-randomization, and applicability domain analysis using the statistical defect approach.

Results

The models exhibited strong predictive performance, with training set coefficients of determination (R²) ranging from 0.8716 to 0.8808 and test set R² peaking at 0.9232. Additional metrics, including Concordance Correlation Coefficient (CCC up to 0.9366), Index of Ideality of Correlation (IIC up to 0.9608), and mean absolute error (MAE as low as 0.1657), confirmed accuracy and generalizability. SMILES-based analysis identified molecular fragments with positive and negative impact on log(M/P.

Conclusion

This study establishes conformation-independent QSPR modeling as an important tool for early-stage drug development and pharmacovigilance, providing actionable insights to minimize infant exposure and improve maternal and infant health outcomes for breastfeeding women through safer therapeutic options. 

Graphical Abstract