Microbial Biomarkers Identification for Human Gut Disease Prediction using Microbial Interaction Network Embedded Deep Learning

Human gut microorganisms are crucial in regulating the immune system. Disruption of healthy relationship between microbiota and epithelial cells leads to development diseases. Inflammatory Bowel Disease (IBD) Colorectal Cancer (CRC) gut-related disorders with complex pathophysiological mechanisms. With massive availability microbiome data, computer-aided microbial biomarker discovery for IBD CRC is becoming common. However, interactions were not considered by many existing identification methods. Hence, this study, we aim construct a interaction network (MIN). The MIN accounts associations formed among microbes hosts. This work explores graph embedding feature selection through construction sparse using MAGMA embedded into deep feedforward neural (DFNN). aims reduce dimensionality select prominent features that form disease biomarkers. selected passed forest classifier prediction. proposed methodology experimentally cross-validated (5-fold) different classifiers, works, models DFNN datasets. Also, biomarkers verified against biological studies highest achieved AUC, accuracy, f1-score 0.863, 0.839, 0.897, respectively, dataset 0.837, 0.768, 0.757, dataset. As observed, method successful selecting subset informative CRC.