Deep Learning-Based Drug Half-Life Classification to Enhance Drug Development and Pharmacokinetics.

Purpose: Predicting drug half-life is essential in pharmacokinetics (PK), influencing dosing strategies and guiding drug development. Traditional regression models estimate exact half-life values but are sensitive to pharmacokinetic variability, limiting their practical use. This study introduces a classification-based approach that separates drugs into short and long half-life groups using a 12-hour threshold, offering clearer clinical interpretability. Methods: Molecular structures were processed using a convolutional neural network (CNN), specifically a fine-tuned AlexNet, to extract high-level features. These extracted features served as inputs for a neural network classifier. A holdout validation strategy was applied, with data split into 70% for training, 15% for validation, and 15% for testing. Model performance was assessed based on classification accuracy and F1-score. Results: The model achieved an F1-score of 90.9% at the optimal feature dimension of 10. Accuracy reached 96.2% on validation data and 92.3% on test data, demonstrating strong generalization capabilities. Compared to regression-based methods, this framework better accounts for variability in drug half-life and yield results that are easier to interpret in clinical contexts. Conclusion: This work proposes an efficient method for drug half-life classification, supporting drug formulation and dosing strategies. The findings highlight the value of classification in early drug development and provide a robust, scalable tool for pharmacokinetic research.