PINNet: a deep neural network with pathway prior knowledge for Alzheimer's disease

Introduction Identification of Alzheimer's Disease (AD)-related transcriptomic signatures from blood is important for early diagnosis the disease. Deep learning techniques are potent classifiers AD diagnosis, but most have been unable to identify biomarkers because their lack interpretability. Methods To address these challenges, we propose a pathway information-based neural network (PINNet) predict patients and analyze brain using an interpretable deep model. PINNet (DNN) model with prior knowledge either Gene Ontology or Kyoto Encyclopedia Genes Genomes databases. Then, backpropagation-based interpretation method was applied reveal essential pathways genes predicting AD. Results The performance compared DNN without pathway. Performances outperformed were similar those gene expressions, respectively. Moreover, considers more AD-related as features than in process. Pathway analysis protein-protein interaction modules highly contributed showed that enriched cell migration, PI3K-Akt, MAPK signaling, apoptosis blood. module included apoptosis, protein ubiquitination, t -cell activation. Discussion By integrating about pathways, can related source codes available at https://github.com/DMCB-GIST/PINNet .