Spatial mapping of subcellular disease pathways and cytoarchitecture to anatomical brain regions for multi‐scale brain simulation with The Virtual Brain

Abstract

AbstractBackgroundComprehensive knowledge bases as NeuroMMSig, a knowledge base of multimodal mechanistic signatures, catalog knowledge about Alzheimer's Disease (AD) and its underlying candidate mechanisms. Besides, digital atlases provide information about cytoarchitectonics and receptor densities, e.g. JuBrain at the EBRAINS platform of the Human Brain Project. However, a gap exists presently with respect to the integration of the various knowledge sources. The Virtual Brain (TVB) is a full brain simulation platform that integrates multimodal imaging data with mathematical brain networks. We present a methodology to map semantic information about AD to 3D brain atlases and to use it for simulations with TVB – increasing the biological plausibility of whole‐brain simulations and disease specificity.MethodOur framework allows topological assignment of brain‐related data from different sources to the Multimodal surface‐based cortical parcellation (MMP) of the Human Connectome Project. Our tool allows mapping of stochastic information from the literature database SCAIView to brain regions. The relevance score for a specific query is calculated for anatomy‐related words from the anatomic ontology Uberon, which we transfer by a unique matrix to MMP parcels. Re‐sampling by the fsaverage brain space of freesurfer allows interoperability with other atlases. In addition, atlas‐annotated data from EBRAINS and neuroimaging data are co‐registered with MMP, enabling knowledge integration into a standardized 3D brain space.ResultOur methodologic pipeline can be used as a software to map semantic information to the brain, which we demonstrate with NeuroMMSig pathways. We further demonstrate how the resulting maps constrain TVB simulations. Face validity is provided through reproducing group‐level (AD versus controls) results of simulations integrating subject‐specific positron emission tomography when semantic information about AD stages is integrated into TVB. We elucidate how brain maps derived from different semantic queries can change simulated brain dynamics fundamentally. Literature‐derived NMDA receptor densities show significant differences between AD and healthy controls. When used to constrain TVB models, this leads to an in silico disruption of functional subnetworks in AD.ConclusionWe developed a tool for mapping semantic knowledge to anatomical brain regions. This enables multi‐scale brain simulations with increases biological plausibility and disease‐specificity for translational research.