Brain volume predicts survival of colliding-spreading messages on mammal brain networks

TL;DR

This study models message congestion in mammal brain networks, revealing that message survival correlates with brain size and is shaped by network topology, with smaller brains supporting longer message persistence.

Contribution

It introduces a colliding-spreading model to simulate message congestion and demonstrates that message survival distribution is an emergent property of network topology and dynamics, independent of physical distances.

Findings

Message survival follows a skewed lognormal-like distribution.

Smaller brains have longer message survival times.

Message survival correlates negatively with brain size (r = -0.64).

Abstract

White matter in mammal brains forms a densely interconnected communication network. Due to high edge density, along with continuous generation and spread of messages, brain networks must contend with congestion, which may limit polysynaptic message survival in complex ways. Here we study congestion with a colliding-spreading model, a synchronous Markovian process where messages arriving coincidentally at a node are deleted, while surviving messages spread to all nearest neighbors. Numerical simulations on a large sample of mammal connectomes demonstrate that message survival follows a positively skewed lognormal-like distribution for all connectomes tested. This distribution mirrors empirical distributions of interareal distances and edge weights. However, the distribution of message survival is an emergent property of system dynamics and graph topology alone; it does not require interareal distances or edge weights. We then show that message survival is well predicted by log brain volume (r = -0.64) across species. Thus, messages survive longer in small compared to large mammals, in accordance with the notion that larger brains become more modular. Chimpanzee showed the lowest message survival among the animals tested. We describe structural properties that may play a role in generating these dynamics and we discuss implications of our results for brain function and evolution.