ProtAlign-ARG: Antibiotic Resistance Gene Characterization Integrating Protein Language Models and Alignment-Based Scoring

Increasing antibiotic resistance poses a severe threat to human health. Detecting and categorizing antibiotic resistance genes (ARGs), genes conferring resistance to antibiotics in sequence data is vital for mitigating the spread of antibiotic resistance. Recently, large protein language models have been used to identify ARGs. Comparatively, these deep learning methods show superior performance in identifying distant related ARGs over traditional alignment-base methods, but poorer performance for ARG classes with limited training data. Here we introduce ProtAlign-ARG, a novel hybrid model combining a pre-trained protein language model and an alignment scoring-based model to identify/classify ARGs. ProtAlign-ARG learns from vast unannotated protein sequences, utilizing raw protein language model embeddings to classify ARGs. In instances where the model lacks confidence, ProtAlign-ARG employs an alignment-based scoring method, incorporating bit scores and e-values to classify ARG drug classes. ProtAlign-ARG demonstrates remarkable accuracy in identifying and classifying ARGs, particularly excelling in recall compared to existing ARG identification and classification tools. We also extend ProtAlign-ARG to predict the functionality and mobility of these genes, highlighting the model’s robustness in various predictive tasks. A comprehensive comparison of ProtAlign-ARG with both the alignment-based scoring model and the pre-trained protein language model clearly shows the superior performance of ProtAlign-ARG. ### Competing Interest Statement The authors have declared no competing interest.