Recent advances in forecasting infectious disease aim ultimately to predict where and when an outbreak may occur, and what wild species will give rise to new human infections. While we continue to improve our ability to predict the timing and locations of recurring infectious diseases, predicting the 'black swans', for example, the emergence of completely novel infectious diseases, or unexpected wild sources is considered an intractable problem, leaving us perpetually reacting to novel disease threats. Recent work has demonstrated the utility of informatics and machine learning tools to increase our capacity to perform this type of prediction, and its promise for building a preemptive strategy to reducing infectious disease.
Using the Zika virus system as an example, we applied a Bayesian machine learning approach to intrinsic biological data of global primate species to reveal what distinguishes the fraction of primate species that are most permissive to Zika virus infection, and therefore most likely to seed repeated spillover events that could lead to long-term enzootic transmission of disease in humans. Our results identified particular primate species that should become immediate targets for surveillance and management for longterm control of Zika virus in the Americas. These analyses also highlight what data deficiencies disproportionately impact our predictive capacity, and underscore the need to continue investing in basic science as a critical foundation for successful forecasting and prediction.