Multilayer network analysis of Alzheimer disease links multiscale paths to the phenotype

Abstract

AbstractBackgroundAlzheimer Disease (AD) cover a wide range of biological scales, from genes and proteins to cells and tissues, up to behavior. However, how these scales, the endotype, interact with one another to define the clinical phenotype remains poorly understood.MethodsHere we conducted a systems biology study of AD, based on multilayer network analysis and deep phenotyping that integrates genomics, CSF biomarkers, tau and amyloid‐beta PET, brain MRI, risk factors and clinical data, obtained from the ADNI collaboration. Multilayer networks were constructed using mutual information, a nonlinear measure of correlation, at each layer and between layers. Boolean simulations identified paths within and among all layers through which dynamic information is transmitted. Multiple linear regression was then used to predict the phenotype (clinical or imaging variables) from paths within the genetic, molecular and imaging layers.Resultsthe most prominent paths that significantly predicted the phenotype included hippocampus atrophy and tau levels from CSF and PET. Our results thus show that integrating the flow of information across biological scales reveals relevant paths for AD.Conclusionpaths identified by multilayer networks can be pursued as biomarkers or new therapeutic targets.