Predicting Antibiotic Resistance from Genomic Sequences Using a Hybrid CNN-RNN Model: A Comprehensive Approach

Antibiotic resistance remains a significant challenge in modern medicine, as bacteria continue to develop mechanisms to evade the effects of antibiotics, leading to increasingly difficult-to-treat infections. This research addresses this issue by employing advanced machine learning techniques, specifically a hybrid model combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), to predict antibiotic resistance using a comprehensive real-world dataset known as DRIAMS. Our hybrid model was evaluated against various existing machine learning approaches, including random forest, decision tree, logistic regression, CNN, and RNN, and demonstrated superior performance with an accuracy of 78.12%, an F1 score of 0.8021, precision of 0.7359, and recall of 0.8815. These findings highlight the model's ability to effectively distinguish between resistant and susceptible genomic sequences, suggesting that it can enhance treatment strategies, reduce dependence on broad-spectrum antibiotics, and lower the risk of treatment failures. Ultimately, this research contributes to improved infection management and public health by providing a robust tool for predicting antibiotic resistance based on genomic data.