Simple Learning Rules Generate Complex Canonical Circuits

TL;DR

This study demonstrates through computational modeling that simple spike-timing dependent plasticity rules can develop cortical microcircuits with canonical inter-laminar connectivity from initially all-to-all connected networks.

Contribution

It shows that basic plasticity mechanisms are sufficient to produce biologically realistic cortical microcircuits from unstructured networks.

Findings

Networks evolve to resemble canonical circuits under balanced plasticity rules.

Enhanced inputs to layer 4 are crucial for proper circuit formation.

The model predicts specific learning computations across cortical layers.

Abstract

Cortical circuits are characterized by exquisitely complex connectivity patterns that emerge during development from undifferentiated networks. The development of these circuits is governed by a combination of precise molecular cues that dictate neuronal identity and location along with activity dependent mechanisms that help establish, refine, and maintain neuronal connectivity. Here we ask whether simple plasticity mechanisms can lead to assembling a cortical microcircuit with canonical inter-laminar connectivity, starting from a network with all-to-all connectivity. The target canonical microcircuit is based on the pattern of connections between cortical layers typically found in multiple cortical areas in rodents, cats and monkeys. We use a computational model as a proof-of-principle to demonstrate that classical and reverse spike-timing dependent plasticity rules lead to a formation of networks that resemble canonical microcircuits. The model converges to biologically reasonable solutions provided that there is a balance between potentiation and depression and enhanced inputs to layer 4, only for a small combination of plasticity rules. The model makes specific testable predictions about the learning computations operant across cortical layers and their dynamic deployment during development.