Revealing the temporal dynamics of antibiotic anomalies in the infant gut microbiome with neural jump ODEs
This work addresses the challenge of anomaly detection in irregularly sampled time-series data, specifically for monitoring antibiotic effects on infant gut microbiomes, offering a translational opportunity to optimize interventions, though it appears incremental as it builds on existing NJODE methods.
The researchers tackled the problem of detecting anomalies in irregularly sampled multivariate time-series data, particularly in data-scarce settings, by developing a framework using neural jump ordinary differential equations (NJODEs). They demonstrated its effectiveness on synthetic data and applied it to infant gut microbiome trajectories, revealing that antibiotic-induced disruptions were more prolonged after second courses, extended treatments, and exposures during the second year of life, with the method outperforming diversity-based baselines in predicting antibiotic events.
Detecting anomalies in irregularly sampled multi-variate time-series is challenging, especially in data-scarce settings. Here we introduce an anomaly detection framework for irregularly sampled time-series that leverages neural jump ordinary differential equations (NJODEs). The method infers conditional mean and variance trajectories in a fully path dependent way and computes anomaly scores. On synthetic data containing jump, drift, diffusion, and noise anomalies, the framework accurately identifies diverse deviations. Applied to infant gut microbiome trajectories, it delineates the magnitude and persistence of antibiotic-induced disruptions: revealing prolonged anomalies after second antibiotic courses, extended duration treatments, and exposures during the second year of life. We further demonstrate the predictive capabilities of the inferred anomaly scores in accurately predicting antibiotic events and outperforming diversity-based baselines. Our approach accommodates unevenly spaced longitudinal observations, adjusts for static and dynamic covariates, and provides a foundation for inferring microbial anomalies induced by perturbations, offering a translational opportunity to optimize intervention regimens by minimizing microbial disruptions.