Deep Transfer Learning Enables Robust Prediction of Antimicrobial Resistance for Novel Antibiotics

Abstract

Antimicrobial resistance (AMR) has become one of the serious global health problems, threatening the effective treatment of a growing number of infections. Machine learning and deep learning show great potential in rapid and accurate AMR predictions. However, a large number of samples for the training of these models is essential. In particular, for novel antibiotics, limited training samples and data imbalance hinder the models’ generalization performance and overall accuracy. We propose a deep transfer learning model that can improve model performance for AMR prediction on small, imbalanced datasets. As our approach relies on transfer learning and secondary mutations, it is also applicable to novel antibiotics and emerging resistances in the future and enables quick diagnostics and personalized treatments.

Artificial Intelligence for Antimicrobial Resistance Prediction: Challenges and Opportunities towards Practical Implementation

“…Therefore, it is important to select models carefully, considering their appropriateness in terms of identification and interpretation. For instance, the authors in [ 38 , 60 , 61 , 62 , 63 , 64 , 65 , 66 ] used deep-learning and machine-learning models to identify different antibiotics. The authors in [ 38 ] used traditional machine learning and CNN to rapidly predict tuberculosis drug resistance accurately from genome sequences.…”

“…Therefore, it is difficult to find insights for the activities of components of the genome against these antibiotics. The authors in [ 66 ] also used the same data, but the objective was to identify new genes based on a deep CNN model. The idea was to train the base model against CIP.…”

“…Therefore, AI models might struggle to learn the underlying mechanism with this type of evolution in resistance. However, deep-learning models, if provided with enough data and designed in a right way, would provide interesting findings [ 30 , 39 , 66 ].…”

“…It is important to select features that are predictive of AMR, but also to avoid including redundant or irrelevant features, as this can negatively impact model performance. Machine-learning models can be prone to overfitting, especially when they are highly complex [ 66 ]. This means that they may perform well on the training data, but poorly on unseen data.…”

MSDeepAMR: antimicrobial resistance prediction based on deep neural networks and transfer learning

“…These pre-trained models can be fine-tuned with small datasets by laboratories with limited sample collection and computing capacity, which can, in such a way, take advantage of powerful models. A recent proposal in this direction is to detect AMR using deep learning using transfer learning based on whole genome sequence data (Ren et al, 2022 ). However, to the best of our knowledge, no transfer learning proposals have been made for AMR based on MS techniques.…”