Purpose: The purpose of this study was to develop a robust and externally predictive in silico QSAR-neural network model for predicting plasma protein binding of drugs. This model aims to enhance drug discovery processes by reducing the need for chemical synthesis and extensive laboratory testing. Methods: A dataset of 277 drugs was used to develop the QSAR-neural network model. The model was constructed using a Filter method to select 55 molecular descriptors. The validation set’s external accuracy was assessed through the predictive squared correlation coefficient Q2 and the root mean squared error (RMSE). Results: The developed QSAR-neural network model demonstrated robustness and good applicability domain. The external accuracy of the validation set was high, with a predictive squared correlation coefficient Q2 of 0.966 and a root mean squared error (RMSE) of 0.063. Comparatively, this model outperformed previously published models in the literature. Conclusion: The study successfully developed an advanced QSAR-neural network model capable of predicting plasma protein binding in human plasma for a diverse set of 277 drugs. This model’s accuracy and robustness make it a valuable tool in drug discovery, potentially reducing the need for resource-intensive chemical synthesis and laboratory testing.
19 October 2025 in Advanced Pharmaceutical Bulletin
Purpose Predicting drug half-life is crucial for pharmacokinetics, impacting dosing strategies and drug development. Traditional regression models estimate exact half-life values but often struggle with pharmacokinetic variability, making direct interpretation and clinical decision-making challenging. This study proposes a classification-based approach that categorizes drugs into short and long half-life classes using a 12-hour threshold, providing a more practical and interpretable alternative. 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 proposed model achieved an F1-score of 90.9% for the optimal feature dimension of 10. The accuracy rates were 96.2% on validation data and 92.3% on test data, demonstrating strong generalization capabilities. The classification framework outperformed regression-based approaches by addressing the inherent variability in drug half-life data and providing a more clinically relevant decision-making tool. Conclusion This study introduces an efficient and interpretable method for drug half-life classification, facilitating improved drug formulation and clinical decision-making. The findings highlight the advantages of classification in early-stage drug development, providing a scalable and robust approach for pharmacokinetic applications.
The emergence of the COVID-19 pandemic has underscored the critical importance of accurate and reliable methods for the early detection and management of cases. Traditional approaches to COVID-19 diagnosis often rely on binary classification methods, which may limit their accuracy and robustness. In this study, we propose a novel approach that leverages chest radiography images for predicting COVID-19 cases. By reframing the classification task as a regression problem, we aim to enhance the accuracy and reliability of our predictive model. Our method involves several key steps. Firstly, we collected a dataset of chest radiography images from confirmed COVID-19 cases and non-COVID-19 cases. Next, we preprocessed the images and extracted relevant features using advanced image processing techniques. We then framed the prediction task as a regression problem, allowing us to model the continuous variation in disease severity rather than relying on binary classification. The predictive model was trained using machine learning algorithms, and both internal and external validation were performed to assess its performance. Our method involves converting the classification task into a regression task, which enables improved accuracy and robustness in the model. We performed both internal and external validation, with R train = 0.91, CV-MSE = 0.0253, and Q = 0.91, indicating high accuracy and reliability in predicting COVID-19 cases. Additionally, we conducted an applicability domain analysis, which showed that 99% of unseen data can be accurately predicted by our model. Our findings suggest that our method can be a valuable tool in the early detection and management of COVID-19 cases, which can ultimately improve patient outcomes and public health. Further validation and testing in real-world clinical settings are needed to confirm the effectiveness and generalizability of our approach.