A Protocadherin Network Model: Mechanisms for Initializing Synaptic Connections

Abstract

The huge diversity of protocadherins (Pdchs) make them strong candidates for molecular codes of individuality for each neuron. The trans binding of protocadherin cis dimers allow for self-recognition through homophilic interaction which signal the cells to repulse, thus playing a role in self avoidance and tiling. However, little research has investigated protocadherins as a mechanism for network formation. This paper aims to investigate the role of differing levels of protocadherin similarity on the functional connectivity of neurons, as well proposing protocadherins as a mechanism for randomization and initialization of synaptic connections through its potential for neuron discrimination. In this paper, two potential mechanism by which Protocadherins could drive synaptic formation were modeled onto an artificial network. In model 1, nodes/neurons that share more isoforms (except when the nodes were exactly the same) were more likely to form a connection. In model 2, the fewer isoforms two neurons/nodes share, the more likely they are to form a connection. The tests for digit discrimination and statistical analysis of these models showed that model 1 and 2 are more effective at learning than a nonrandom baseline model in which weights were all connected equally. However, they are less effective at learning than the random baseline model in which weights were assigned from a gaussian distribution. Between the two protocadherin models, model 2 performed better, indicating that protocadherin-driven synapse formation may occur through a mechanism in which dissimilarity is favored.