Synaptic plasticity rules used in current computational models of learning are generally insensitive to physiological factors such as spine voltage, animal age, extracellular fluid composition, and body temperature, limiting their predictive power. Here, we built a biophysically detailed synapse model inclusive of electrical dynamics, calcium-dependent signaling via CaMKII and Calcineurin (CaN) activities. The model combined multi-timescale variables, milliseconds to minutes, and intrinsic noise from stochastic ion channel gat- ing. Analysis of the trajectories of joint CaMKII and CaN activities yielded an interpretable geometrical readout that fitted the synaptic plasticity outcomes of nine published ex vivo experiments covering various spike-timing and frequency-dependent plasticity induction protocols, animal ages, and experimental conditions. Using this new approach, we then generated maps predicting plasticity outcomes across the space of these stimulation conditions. Finally, we tested the model’s robustness to in vivo-like spike time irregularity, showing that it significantly alters plasticity outcomes.
Reference
@article{Rodrigues2021a,
author = "Y. E. Rodrigues and H. Marie and C. O’Donnell and R. Veltz",
title = "A stochastic model of hippocampal synaptic plasticity with geometrical readout of enzyme dynamics",
year = 2023,
journal = "Elife",
month = "Aug",
url = "https://elifesciences.org/articles/80152"
}
