Mostly Biomathematics Lunchtime Seminar
A Computational Model on Cellular-Molecular Description of Early Long-Term Potentiation – Toward “One-Shot” Learning
Speaker: Guanchun Li, Courant Institute of Mathematical Sciences
Location: Warren Weaver Hall 1314
Date: Tuesday, November 22, 2022, 12:30 p.m.
Synopsis:
Synaptic plasticity [the long-term potentiation/depression (LTP/D) of the strength of the synapses that connect neurons in the brain] underlies learning and memory – as described by “neurons that fire together, wire together”. One remarkable type of learning, termed “one-shot learning”, is exemplified by the hippocampal CA1 “place cells” of rodents that learn spatial location with only one (or a very few) passes through the spatial location being learned. Currently, the cellular-molecular mechanisms of LTP/D are not thoroughly understood, especially those responsible for “one-shot” learning.
This work is focused on modeling the molecular and cellular mechanisms of long-term potentiation for CA1 pyramidal neurons in the hippocampus. We use a deterministic ODE representation to model the biochemical reactions of early LTP/D in a single compartment of the dendritic spine head at a synapse, containing well-mixed Calcium ions, Calmodulin, kinase (CaMKII), and phosphatase. When active (phosphorylated) kinase dominates, LTP occurs; when active (dephosphorylated) phosphatase dominates; LTD. The computational model leads to a tri-stable system for the temporal profiles of these quantities, representing the three states (LTP, LTD, and Basal). Under two distinct classes of stimulation, the model produces two distinct learning rules – well known Hebbian STDP (spike timing dependent plasticity), and a recently discovered [Bittner, et al (2017)] learning rule BTSP (behavioral timescale plasticity) that is likely to underlie “one-shot learning”. Within the model, we study and contrast the properties and
mechanisms of STDP and BTSP. In addition, the model also shows a possible role of CaV1 calcium channels in producing a “priming mechanism” that could enhance both BTSP and “one-shot” learning.