Publication

Abstract : Antimicrobial resistance (AMR) is an urgent and pressing global health challenge, and requires innovative methods for its prediction and management for better treatment plans for gonorrhea. Machine learning models are developed for predicting AMR in bacteria based on whole-genome sequence and clinical sample data. The ML methods extrapolate learned pattern occurrences in clinical samples and genomic data to infer potential resistance in a pathogen to a given antibiotic. The models used for the AMR prediction include Support Vector Machine (SVM), and neural networks like Convolution NN, Feedforward NN, and Recurrent NN. The review touches on the challenges of microbial genome diversity and the interpretability of machine learning models by visualizing data like average resistance level by country, the most resistant anitbiotic by country, and Spearman and Pearson correlation. The results have variable data quality, feature selection, and model optimization strategy parameters, such as hyperparameter tuning, which affect the accuracy in predicting AMR. Out of the ML models used SVM has the highest accuracy in terms of AMR prediction. It is the promise and use of ML-based AMR prediction systems in clinical practices to provide knowledge-guided personalized options to aid AR and improve the health outcomes of patients.