Multiscale networks in Alzheimer's disease identify brain hypometabolism as central across biological scales.

Alzheimer's disease encompasses multiple biological scales, spanning molecular factors, cells, tissues, and behavioral manifestations. The interplay among these scales in shaping the clinical phenotype is not yet fully comprehended. In particular, there is great interest in understanding the heterogeneity of the clinical aspects of AD in order to improve treatment and prevention, by targeting those aspects most susceptible to the disease. Here we employed a systems biology approach to address this issue, utilizing multilayer network analysis and deep phenotyping. This integrative analysis incorporated genomics, cerebrospinal fluid biomarkers, tau and amyloid beta (Aβ) PET imaging, brain MRI data, risk factors, and clinical information (cognitive tests scores, Clinical Dementia Rating and clinical diagnosis) obtained through the ADNI collaboration. Multilayer networks were built based on mutual information between the elements of each layer and between layers. Boolean simulations allowed us to identify paths that transmit dynamic information across layers. The most prominent path for predicting variables in the cognitive phenotype layer included the PET radiotracer fluorodeoxyglucose (FDG) in the posterior cingulate. Combinations of different symptomatic variables, mainly related to mental health (depression, mood swings, drowsiness) and vascular features (hypertension, cardiovascular history), were also part of the paths explaining the average phenotype. Our results show that integrating the flow of information across biological scales reveals relevant paths for AD, which can be subsequently explored as potential biomarkers or therapeutic targets. In particular, our results point for paths related with brain hypometabolism as a key feature in AD.