In the presentation, we talked about how machine learning can be used for the discovery of novel antimicrobial resistance genes. Machine learning models trained with known resistance genes or pangenomes reached the highest performances. However, the pangenome model does not lose any robustness when the known resistance genes were taken out from the pangenomes. These results indicate that we have room for the discovery of novel antimicrobial resistance genes. The candidate antimicrobial resistance genes detected in this study should be tested using in vitro techniques to validate the accuracy and interpretability of machine learning models.
Derya Aytan-Aktug’s presentation