Predicting population dynamics of antimicrobial resistance using mechanistic modeling and machine learning

Abstract

Antimicrobial resistance (AMR) infections have become a global public health burden. The pipeline for new antibiotic discovery is draining due to the rapid emergence of resistance to new antibiotics, the limited economic return, and regulatory hurdles. Current strategies to combat the AMR crisis include improving clinical practices under antibiotic stewardship and repurposing FDA-approved drugs. Quantitative modeling of the population dynamics of AMR can inform these strategies by identifying key mechanisms and consequences of resistance development and predicting resistance persistence, with the potential of guiding treatment design. Here we review the current progress of using mechanistic and machine learning (ML) models to understand and predict the population dynamics of AMR in microbial communities. We highlight the current challenges in mechanistic model construction, explore how ML can overcome these limitations, and discuss the translational potential of the computational models.