ProtGAT: Refining Antimicrobial Resistance Detection with Graph-Based Deep Learning*
Anushka Naik, Akanksha Naik, Pratiksha Deshmukh · 2025
The rise of antimicrobial resistance (AMR) is a critical threat to global health, necessitating urgent advancements in diagnostic capabilities. Current methods for detecting AMR are hampered by slow processing times and a lack of precision, making early and accurate detection challenging. This paper introduces ProtGAT, a novel computational framework that combines ProteinBERT and Graph Attention Networks (GAT) to enhance the detection of AMR. Our model leverages deep learning to analyze complex protein sequences and their interactions, improving the accuracy of AMR predictions. The integration of ProteinBERT provides a robust feature extraction from protein sequences, while GAT focuses on identifying key relational features within these sequences. Early testing of ProtGAT demonstrates its superior performance over traditional models, particularly in identifying novel AMR sequences that evade conventional detection methods. This study highlights the potential of advanced computational approaches to revolutionize AMR diagnostics, offering faster and more reliable tools for healthcare providers in combating this growing public health concern.