Long-term plasticity induces sparse and specific synaptic changes in a biophysically detailed cortical model

Abstract

Synaptic plasticity underlies the brain’s ability to learn and adapt. While in vitro experiments reveal the mechanisms behind plasticity at the level of individual pairs of neurons, they lack the scale to explain how they are coordinated in microcircuits to achieve learning. Conversely, research at the population level still relies on in silico approaches of limited generalizability. To overcome these limitations, we embedded a calcium-based model of functional plasticity that captures the diversity of excitatory connections in a thoroughly validated large-scale cortical network model and studied how plasticity shapes stimulus representations at the microcircuit level. We then used an openly available electron microscopic reconstruction of cortical tissue to confirm our testable predictions. We found that in an in vivo-like network state, plasticity acted sparsely and specifically, keeping the firing rate stable without additional homeostatic mechanisms. Our results predict that this specificity at the circuit level is governed by co-firing functional assemblies, spatial clustering of synapses on dendrites, and the topology of the whole network’s connectivity. These effects cannot be captured with point neuron models, random connectivity and pairwise rules. In summary, our findings elevate descriptions of plasticity rules to the population level, bridging the scales between plasticity and learning in networks.