Perturbation-Aware Neural ODE (pNODE) Learns Microbiome Dynamics from Clinical Data and Predicts Gut-Borne Bloodstream Infections in Patients Receiving Cancer Treatment

Abstract

Disruption of the gut microbiota during cancer treatment, particularly in allogeneic hematopoietic cell transplantation (allo-HCT), contributes to adverse clinical outcomes, including gut-borne bloodstream infections. Accurately forecasting microbial population dynamics under clinical perturbations, such as antibiotic administration, could inform treatment strategies to reduce infection risk. However, traditional models like the Generalized Lotka-Volterra (gLV), which consider only pairwise interactions with constant sign and magnitude, are limited in capturing the real-world nonlinear dynamics of multispecies microbiomes following ecosystem disturbances. Here, we introduce a perturbation-augmented Neural Ordinary Differential Equation (pNODE) framework that flexibly models microbial population dynamics in continuous time, integrating both microbial abundances and time-resolved antibiotic perturbations. Using synthetic and real clinical data from over 1,000 allo-HCT patients, we show that pNODEs outperform gLV in predictive accuracy, robustness to noise, and forecasting critical events, such as the intestinal expansion of an opportunistic pathogen. Notably, we demonstrate that running a pre-trained pNODE in generative mode to simulate prospective Enterococcus abundance trajectories from an initial sample and antibiotic timeline yields scores that accurately predict subsequent bloodstream infections in held-out cohorts, outperforming baseline and ground-truth-based predictors. Our findings demonstrate the potential of pNODEs as a next-generation tool for modeling clinical microbiome dynamics, with applications for predicting infections in immunocompromised patients hospitalized for cancer treatment.