Hybrid CNN-RNN models for accelerating antiviral drug discovery: A computational framework for virus-drug predictions
Nagamani H Shahapure, Ravi Babu Birudugadda, Ankur Singh Bist, Ch Srivardhan Kumar, Allimalli Durga Bhavani · 2025
The ongoing challenge of managing viral infections calls for swift and effective antiviral therapy discovery. Traditional drug discovery, often lengthy and resource-intensive, underscores the need for computational models that can expedite drug repurposing. This study presents a novel hybrid deep learning framework, combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), designed to predict virus-drug associations and accelerate antiviral repurposing efforts. The model utilizes CNNs for capturing spatial and hierarchical features and RNNs for recognizing temporal dependencies within virus-drug interactions. Trained on a comprehensive dataset including FDA-approved drugs and data from the DrugVirus2 database, the model achieved substantial performance metrics, including an accuracy of [X%], AUC of [Y], and F1 score of [Z], demonstrating notable improvements over baseline models. These results validate the model&s;s robustness and scalability, highlighting its potential in real-world applications. Future work includes data augmentation and hyperparameter tuning to further enhance adaptability to emerging viral threats. This framework offers a scalable solution crucial for rapid antiviral drug discovery and public health preparedness.