Using Autoencoders for Predicting Latent Microbiome Community Shifts Responding to Dietary Changes

摘要

Human gut microbiome responds rapidly to diet and has important consequences to host health such as obesity and other metabolic disease. Computational tools of modeling microbiome dynamics responding to diet are import in facilitating the development of effective prebiotic and probiotics for disease prevention and treatment. In this work, we present a deep neural network to predict gut microbiome responding to diet change. Our model uses an autoencoder to capture intrinsic structure in data and an artificial neural network to model the nonlinear dynamics of microbiome. Our evaluation using three time-course datasets of murine gut microbiome shows promising results.