Bridging Neural Scales in Epilepsy Research for Human Health: A Review and Conceptual Framework for Multiscale Computational Modeling

Abstract

Epilepsy is a neurological disorder caused by recurrent seizures due to a hypersynchronous activity of neurons that can spread out to larger networks. How this activity on smaller scales transfers to higher levels is hard to capture computationally.   Previous works usually focus on either microcircuit level biophysics or large-scale network analysis with seizure prediction. Here we review the current state of computational           epilepsy research with a specific focus on hippocampal microcircuits, connectome-based models, neural mass and spiking network analyses, and machine learning-based seizure detection and prediction. We highlight a general lack of mechanistic understanding of how cellular-level pathophysiology contributes to altered network function across spatial and temporal scales. Using the example of the Neurobridge, we propose a conceptual framework that connects a computationally reconstructed biophysically detailed hippocampal microcircuit with a connectome-based model of the whole-body C. elegans nervous                                                                                                                                                                                                  system using Brain2 and NeuroML-based standardized descriptions and the c302         modeling environment. We discuss how such a framework could be utilized to understand the effects of seizures on a systems level and how these could affect emergent behaviors. We review common evaluation schemes and benchmarking tools for microcircuit,           network, machine learning, and closed-loop epilepsy analysis at different scales. Finally, we discuss how such frameworks could be utilized for therapeutic modulation, drug        discovery, neuromodulation, and personalized medicine approaches in epilepsy research. Our analysis highlights that only a minor fraction of existing epilepsy computational      studies tackle the challenges of multiscale analysis and emphasize differences in              conceptual modeling, parameterization, simulation environments, and general                     reproducibility between different domains. We argue that Neurobridge provides a             conceptual basis for future computational epilepsy studies that aim to be maximally        multiscale and discuss possible future directions including efficient co-simulation and      experimental validation, learning-based graph analysis, and similar approaches for other neurological diseases including Parkinson’s or Alzheimer’s disease.