Background: Neuropsychological and neurophysiological analyses focus on understanding how neuronal activity and co-activity predict behavior. Experimental techniques allow for modulation of neuronal activity, but do not control neuronal ensemble spatiotemporal firing patterns, and there are few, if any, sophisticated in silico techniques which accurately reconstruct physiological neural spike trains and behavior using unit co-activity as an input parameter.
New method: Our approach to simulation of neuronal spike trains is based on using state space modeling to estimate a weighted graph of interaction strengths between pairs of neurons along with separate estimations of spiking threshold voltage and neuronal membrane leakage. These parameters allow us to tune a biophysical model which is then employed to accurately reconstruct spike trains from freely behaving animals and then use these spike trains to estimate an animal's spatial behavior. The reconstructed spatial behavior allows us to confirm the same information is present in both the recorded and simulated spike trains.
Results: Our method reconstructs spike trains (98 ± 0.0013% like original spike trains, mean ± SEM) and animal position (9.468 ± 0.240 cm, mean ± SEM) with high fidelity.
Comparison with existing method(s): To our knowledge, this is the first method that uses empirically derived network connectivity to constrain biophysical parameters and predict spatial behavior. Together, these methods allow in silico quantification of the contribution of specific unit activity and co-activity to animal spatial behavior.
Conclusions: Our novel approach provides a flexible, robust in silico technique for determining the contribution of specific neuronal activity and co-activity to spatial behavior.
Keywords: Biophysical simulation; Neural spike trains.
Copyright © 2022 Elsevier B.V. All rights reserved.