A Generative Model for Cortical Networks

Speaker: Dr. Daniel L. Barabasi , University of Notre Dame, Department of Physics

Date: Tuesday, May 23, 2017

Time: 3:00 PM to 4:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: Star D463

Event Type:

Room Description:

Host: Manolis Kellis, Computation & Biology Group at CSAIL, MIT

Contact: Geraldine McGowan, gmcgowan@csail.mit.edu

Relevant URL:

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: A Generative Model for Cortical Networks

Updated experimental efforts using hemisphere-wide retrograde and anterograde tracing have provided large-scale static data about the architecture of the cortex. Previous studies of low-density interareal connectivity have found emergent scale-free and rich club properties as indicators of network heterogeneity, however these need to be revised due to the high-density nature of the cortico-cortico wiring reported in the recent experimental studies. To understand the level of specificity and connection heterogeneity of the cortical network one needs to take an alternative approach that is more suitable for dense, directed, spatial networks. Past work has introduced a simple core-periphery detection method based on clique distribution analysis, showing a strong core-periphery organization of the cortex in both macaques and mice when compared to appropriately chosen null-models, which was further validated using stochastic block modeling analysis. Physically, tracing experiments have revealed in both macaque and mouse brains the action of the so-called Exponential Distance Rule (EDR): an exponential decay of connection probability with distance. A prior network model based on the EDR was defined at the level of functional areas and captured many features of the experimentally obtained network, however, as we show here it fails in other measures, such as in identifying the core-periphery structure, found in experiments. Here we introduce a neuronal-level network model based on the EDR, which, given a a parcellation of the cortex into functional areas, naturally generates the interareal network. This new model now also quantitatively captures the features missed by the areal-level model and shows that the core-periphery structure emerges due both to the EDR and the distribution of area sizes and their relative positioning. These findings demonstrate that the cortical network cannot be modeled as a simple connectivity graph at the areal level, but one needs multiscale network models that are intricately tied with the functional parcellation of the cortex and the morphological properties of the cortical plate.

Research Areas:

Impact Areas:

This event is not part of a series.

Created by Deborah Goodwin at Thursday, May 18, 2017 at 11:46 AM.