The mammalian brain is believed to be optimally designed for robust and adaptable computation of the
sensory inputs from the world, with respect to both its hardware (network structure) and software (network
dynamics). The precise connections between the intricate structural connectivity and the rich network
dynamics, however, are yet unknown. Moreover, our understanding of how the network structure and
dynamics shape (or are shaped by) underlying coding principles in the brain network, is limited. My research
plan proposes to close this gap by leveraging rich dataset obtained by state-of-art experimental techniques at
the Allen Institute for Brain Science and innovative mathematical methods.
Specifically, my project aims to 1) link network structure and dynamic information processing in the brain,
and to 2) bridge the gap between detailed biophysiological mechanisms and overarching neural coding
principles with a focus on predictive coding theory, using data-driven mathematical models. To address Aim 1,
I will investigate how network dynamics measured by synchronizability, metastability, and integrated
information depend on local and global structure of the network. I will then study whether the experimentally
obtained mouse brain connectome has optimal connectivity structures for unique dynamical characteristics.
These analyses will be extended to the cell-type and layer-specific brain connectivity, based on the latest Allen
Mouse Brain Connectivity data obtained from Cre-transgenic mice. During the independent phase, I will
investigate whether brain-like networks can be evolved from optimization of dynamic measures. Regarding Aim
2, I will analyze data obtained from my current collaborative project which experimentally tests predictive
coding models in the mouse visual cortex. In this study, we measure neural activity in response to expected
and unexpected sequences of natural stimuli across three hierarchically related areas. Upon completion of the
experiments, during the mentored phase, I will investigate mapping of algorithmic units in predictive coding
models to neuronal populations in different layers. During my independent career period, I will extend the
predictive coding model to incorporate active sensing and thalamo-cortical circuitries.
The project during the mentored phase will be carried out at the University of Washington which provides a
highly interdisciplinary environment and offers the ideal training for me to become an independent researcher. I
will also have access to rich resources and outstanding collaborators at the Allen Institute for Brain Science. I
will have two mentors, one from the University of Washington and another from the Allen Institute for Brain
Science. This unique setup will allow me to study mathematical models based on experimental data obtained
by cutting-edge techniques with guidance from mentors with strong theoretical backgrounds. With theories
closely tied to experiments, I believe my proposed project will contribute to our understanding of the connection
between structure and computation of the neuronal network, addressing BRAIN initiative’s high priorities.