30.03.2017: xiao-jing wang

What does it mean to build a large-scale brain circuit model?

In this talk, I will first promote the idea of variations on the theme of a canonical local circuit, contrasting sensory microcircuits dedicated to coding and processing stimuli from senses with "cognitive-type" microcircuits capable of working memory and decision-making. This line of research has led us to study multi-region brain systems based on mesoscopic connectome and physiological experiments. We have developed large-scale cortex modeling of macaque monkey and mouse. By taking into account variations across cortical areas, the model naturally gives rise to a hierarchy of timescales and, endowed with a laminar structure of the cortex, it captures frequency-dependent interactions between bottom-up and top-down processes. Moreover, in a complex brain system, routing of information between areas must be flexibly gated according to behavioral demands. We proposed such a gating mechanism with a disinhibitory circuit motif implemented by three subtypes of (PV+, SOM+ and VIP+) inhibitory neurons, and I will report a recent finding that the relative distribution of these three interneuron classes varies markedly across the whole mouse cortex. Finally, I will show preliminary results on distributed working memory representation. Circuit modeling across levels, combined with training multi-module recurrent networks, represents a promising approach to elucidate high-dimensional dynamics and functions of the global brain.