CaDS Seminar 2026 - May 19

Sinovia Fotiadou (SDL Neuroscience)

Bridging NEST Microcircuits to Zerlaut Mean-Field Models: A Reproducible Single-Region Workflow for the Virtual Brain Twin

Abstract:

This talk presents a reproducible workflow bridging microscopic spiking neural network simulations and mesoscopic neural mass models. It is motivated as a methodological building block in the Virtual Brain Twin project, where altered excitation–inhibition balance is a relevant hypothesis in schizophrenia research. Here we use transformer models to create robust dynamic maps between the activity at the microscopic and the mesoscopic scales of the brain. The study asks how far a conductance-based AdEx spiking neural microcircuit simulated in NEST can be linked to an AdEx mean-field model in the style of Zierlaut/ di Volo under identical time-varying drive, and which practical failure modes dominate in an operational pipeline.