Computational Principles of Brain Network Development and Dynamics

Understanding the development and dynamics of brain networks requires integrating empirical data and theoretical models across multiple biological scales, from molecular and genetic networks to cortical neural circuits. I will discuss two of my research projects combining novel computational and theoretical modeling tools with large-scale empirical data sets to investigate how neurons and glia acquire their identities, and how their activity gives rise to global brain states. In the first part, I will examine how networks of epigenetic elements shape the development and mature identity of brain cells. Applying bioinformatics and machine learning tools to whole-genome bisulfite sequencing (WGBS) data, we discovered novel patterns of DNA methylation that are reshaped during brain development and which distinguish brain cell types. Throughout the genome, we find that methylation is closely linked to the function of genes and regulatory sequence elements with specific roles in neuronal and glial cell function. By combining neurobiological analysis with new computational methods for large-scale epigenomics data, our research provides a framework for understanding the genetic regulatory mechanisms of normal brain cell differentiation and points the way to future studies of epigenetic disregulation in neurological and psychiatric disease.

In the second part of the seminar, I will explore the role that oscillatory neural activity across a range of low-frequency bands plays in disrupting information processing and producing unconsciousness during general anesthesia. To address this in human subjects, we have combined high-density EEG and intracranial ECoG and depth electrode recordings, together with biophysical model-based source localization and statistical signal processing techniques. Because conventional power spectral analyses treat each frequency band independently, we use higher-order statistics to integrate information about neural activity across the EEG spectrum. Following loss of consciousness we observe a novel transition between global brain states with distinct patterns of phase-amplitude cross-frequency coupling. Our data suggests that general anesthesia induced unconsciousness is not a unitary state of the cortical network; rather, cortical dynamics may be organized in multiple distinct modes depending on drug concentration. Phase-amplitude coupling may be a useful supplement to the power spectrum and bispectrum-derived measures in clinical EEG-based monitoring of brain state.