A multi-label learning framework for predicting antibiotic resistance genes via dual-view modeling

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Antibiotic resistance is a major global health crisis, necessitating accurate identification of antibiotic resistance genes (ARGs) for safety and clinical management.
• Traditional culture-based methods for ARG identification are accurate but slow and require expertise.
• Current computational methods using whole genome sequencing and sequence similarity have limitations, including failure to detect ARGs with low sequence identity and ineffective multidrug resistance prediction.

Purpose of the Study:

• To develop an advanced computational framework for predicting antibiotic resistance genes (ARGs) in bacteria.
• To address the challenges of low sequence identity detection and multidrug resistance prediction in existing ARG identification methods.
• To improve the accuracy and efficiency of ARG identification for enhanced safety regulation and clinical decision-making.

Main Methods:

• Proposed an end-to-end multi-label learning framework for ARG prediction.
• Modeled ARG prediction as a multi-label learning problem using a deep neural network.
• Incorporated a specialized loss function to leverage multi-label learning advantages.
• Employed a dual-view modeling mechanism integrating sequence-based and structure-based ARG information.

Main Results:

• The proposed deep neural network framework demonstrated effectiveness in predicting ARGs.
• The multi-label learning approach and dual-view mechanism improved prediction accuracy.
• Experimental results on publicly available data validated the framework's performance.

Conclusions:

• The developed end-to-end multi-label learning framework offers a robust solution for predicting antibiotic resistance genes (ARGs).
• This approach effectively addresses limitations of existing methods, particularly in detecting novel ARGs and predicting multidrug resistance.
• The findings contribute to advancing bacterial safety regulation and clinical treatment strategies against antibiotic resistance.