Prediction of milk/plasma drug concentration (M/P) ratio using support vector machine (SVM) method

Area of Science:

• Pharmacokinetics
• Computational Chemistry
• Drug Safety

Background:

• Predicting milk-to-plasma (M/P) drug concentration ratios is crucial for infant safety.
• Existing computational models face challenges in accurately classifying drug transfer.
• Identifying drugs with high M/P ratios is essential to mitigate risks for nursing infants.

Purpose of the Study:

• To develop a reliable computational model for predicting/classifying the milk-to-plasma (M/P) drug concentration ratio.
• To utilize the Support Vector Machine (SVM) algorithm to assess the potential risk of drugs to nursing infants.
• To compare the efficacy of SVM with other classification methods for M/P ratio prediction.

Main Methods:

• Drugs were characterized using a comprehensive set of molecular descriptors.
• Five key molecular descriptors were identified as most important for predictive model construction.
• Two classification models, linear discriminant analysis (LDA) and SVM, were developed and validated using bootstrapping.

Main Results:

• The SVM model achieved high classification accuracy: 90.63% for the training set and 90.00% for the test set.
• The overall accuracy of the SVM model was 90.48%, significantly higher than LDA's 77.78%.
• SVM demonstrated superior performance compared to LDA in classifying drug M/P ratios.

Conclusions:

• The Support Vector Machine (SVM) method is an effective tool for evaluating drug risks in nursing infants.
• SVM provides a reliable approach for predicting M/P ratios when experimental data is unavailable.
• This computational approach aids in drug safety assessments for breastfeeding populations.