Machine learning methods to decipher the evolutionary dynamics of microbial pathogens.

Machine learning is increasingly enabling predictive analyses of microbial pathogen evolution, including adaptation and drug resistance, using genomic data paired with geographic and risk-group metadata.

• Why it matters: The field faces scalability bottlenecks from heterogeneous, high-dimensional pathogen and epidemic datasets, limiting retrospective monitoring and constraining proactive forecasting.
• What they did: The review synthesizes ML applications that use genomic sequences with metadata (e.g., geographic locations, risk groups, sampling information) to predict adaptation and drug resistance, infer sources and hosts, flag high-risk lineages, and estimate phylodynamic and epidemiological parameters.
• The result: These methods are shifting molecular epidemiology toward a proactive, predictive framework for major global microbial threats, while also highlighting limitations in interpretability, uncertainty quantification, data-volume needs, and sensitivity to model misspecification.

AI summary of the abstract · may contain errors · not medical advice · read the paper.