Characterizing metabolic activities in microbial communities is crucial for understanding the functions of microbes and their impact on the host or environment. Computational methods that are capable of predicting metabolomic profiles from microbial compositions save considerable efforts needed for metabolomic profiling experimentally. Numerous computational methods have been developed to solve this problem. Yet we still lack a method with high prediction power and general applicability. Here we develop a new method, mNODE (Metabolomic profile prediction using Neural Ordinary Differential Equations), leveraging a new family of deep neural network models. We show compelling evidence that mNODE outperforms existing methods. Moreover, it can naturally incorporate dietary information to further enhance the prediction of metabolomic profiles. The presented results demonstrate that mNODE is a powerful tool to investigate the microbiome-diet-metabolome relationship, facilitating future research on precision nutrition.