Todd W. Troyer - US grants

Recent studies relating dyslexia to temporal processing highlight the importance of temporal tasks for learning and cognition. Previous computational modeling of vocal learning in song birds suggests that changes in neural excitability on the time scale of single vocal gestures may be an important mechanism by which motor programs are modified to match an internally memorized sensory template. The research proposed here will use the in vitro brain slice preparation to investigate the ability of N- methyl-D-aspartate receptors to increase neuronal excitability, providing a transient facilitation to subsequent synaptic input. By characterizing basic neural mechanisms in the context of a functional model, the proposed investigation will further our understanding of the neural basis of complex imitation-from- memory tasks, especially that of human speech development.

DESCRIPTION (provided by applicant): The combination of an ethologically important, stereotyped behavior and specialized anatomy make avian song learning an ideal system in which to study the neural basis of motor learning. Auditory feedback from a bird's own vocal output is crucial for song acquisition, and song learning shares many important similarities with human speech learning. Furthermore, basal ganglia circuits that play an crucial role in song learning contribute during motor learning in mammals and have been implicated in human motor disorders such as Parkinson's disease. A database of songs will be constructed by recording from a large number of developing songbirds. All vocalizations produced during the period of song learning will be recorded, enabling theexamination of a complex learned behavior with unparalleled detail and completeness. Advanced statistical techniques will be used to analyze the detailed time course of song development, focusing on possible interactions between the development of individual song "syllables," song sequence, and song rhythm. A solid behavioral database will facilitate the interpretation of a wide range of experiments in the field, and will be used to develop more comprehensive models of song learning. By clarifying the principles governing vocal learning in songbirds, these models will yield insight into fundamental questions related to motor learning in humans.

The proposed research will initiate a cross-institutional collaboration between an electrophysiologist and a computational neuroscientist to examine fundamental questions about how the brain produces complex sequences of behaviors. Research will be conducted using songbirds, animals that produce a richly structured sequence of highly reproducible song syllables. Previous studies have recorded the activity individual brain cells while birds are singing, and have shown that neural activity is locked to song production at a temporal scale of several thousandths of a second, a level of precision rarely achieved in the study of complex natural behaviors. This proposal will exploit this precise relationship between brain activity and sequential behavior by recording from individual birds as they sing many song renditions. Sophisticated analysis tools will then be used to examine both brain activity and song output in great detail. Subtle variations in syllable features and syllable sequencing will be used as natural experiments to determine how the precise activity of individual nerve cells are grouped together to form behavioral units (the syllables), and how these groupings of neural activity are strung together into syllable sequences. Guided by these data, computer models will be constructed to better understand the underlying biological mechanisms that orchestrate complex sequences of brain activity. As part of the proposed interdisciplinary research, novel algorithms for the fine-grained analysis of vocal behavior will be developed, and these may find application in related fields ranging from motor control to robotics to human speech. The proposal also supports a cross-institutional student exchange between the University of Texas at San Antonio, a Hispanic-Serving Institution, and the University of California San Francisco, a world-renowned center for biomedical research.

The objective of this EAGER project is to build and apply a computational toolbox to study and model power-law dynamics in the brain. Traditionally, any complex behavior in neuroscience is broken into the interactions of multiple components, each working in its own characteristic temporal framework. However, there is an increasing number of examples, such as in brain activity recording by electroencephalography (EEG), firing rate adaptation, and synaptic weight dynamics, in which the characteristic process follows power-law dynamics, which indicate that the time constant of a mechanism at one scale is highly correlated to the activity of the system at multiple scales. Therefore, the overall behavior of the system cannot be separated into largely independent components and traditional analysis techniques cannot provide an appropriate description of how the system works. In order to understand neuronal information processing at multiple scales it is necessary to develop a framework to analyze and model power-law dynamics at all levels of biological organization. This project plans to make widely available a unified platform to detect, analyze, validate, and model power-law behavior in the nervous system at multiple scales of organization. To broaden impact the team will generate products for the public that will explain the differences between power-law and exponential processes and their importance in neuroscience research. Research opportunities will be provided for students, especially underrepresented group at the University of Texas at San Antonio (UTSA), a minority serving institution.

The collaborative team will analyze and model power-law relationships in large-scale brain activity and complex behavior. The project aims to build and validate a toolbox to test and characterize power-laws in data streams and to model power-law dynamical systems. For this purpose state-of-the art algorithms will be used to characterize experimental data and fractional differential equations to model power-law dynamical systems. This modeling platform will allow the study of power-law processes from the sub-cellular to the behavior scales. The toolbox will be applied to two very different problems dealing with complex pattern generation (birdsong production) and human language comprehension. Both applications will require the analysis of Big Data streams and model non-linear sequence production or decision-making. Although initially the focus of research will be in these two projects the framework will be built to be applicable to a wide range of neuroscience projects that can impact the research done under the BRAIN initiative. Interactive examples will be implemented using Mathematica and Matlab platforms and the team will also update and write new Wikipedia pages on the topics of this grant. In all projects, graduate and undergraduate students will be involved in both the research and educational components, providing opportunities not only to do research but to enhance their communication skills.