An anatomically and biophysically realistic rodent subthalamic nucleus neuron model

Abstract

The subthalamic nucleus (STN) is the primary target for deep brain stimulation (DBS) to treat the motor symptoms of Parkinson's disease (PD). However, the mechanisms of action of STN DBS remain unclear. Computational models of STN neurons are used to investigate the mechanisms of STN DBS and study the biophysics of STN local field potentials (LFPs). However, the only existing multicompartment model of an STN neuron, Gillies and Willshaw (GW) model, is anatomically and biophysically incomplete. It lacks an axon and has an unrealistic ion channel distribution. Therefore, we improved upon the GW STN neuron model and optimized its biophysics (21 biophysical parameters) using experimental electrophysiological recordings as validation constraints. Parameter optimization was performed with a genetic algorithm which generated around half a million phenotypes using crossover and mutation. Then it selected the best candidates employing several cost functions. These cost functions were based on electrophysiological features commonly used to characterize a neuron and include spontaneous spike rate, membrane potential in response to a hyperpolarization current, and the frequency-current curve. The costs for each phenotype were summed to select the best performing model. The resulting STN neuron model performed better than the original GW model and can be used in computational modeling of STN DBS and LFPs.