John M. Beggs - US grants

The physical sciences have had great success in describing how complex phenomena can emerge from the collective interactions of many similar units. Waves, synchrony, phase transitions, and self-organization are all examples of this. Although the brain is tremendously complex, it is composed of many units, neurons, which appear to be similar. This resemblance has led many researchers to borrow concepts from physics in an effort to explain neural function. Simulations indicate that stable states can be used to store information, and that the critical point maximizes both information transmission and information storage. While this body of theory has prospered, experiments to test it have been few. New advances in technology have allowed thousands of interconnected neurons to be grown in culture on microfabricated arrays of many electrodes. These cultured brain slices can be kept alive for weeks while their spontaneous electrical activity is recorded. The large data sets produced by these experiments allow many of the hypotheses inspired by statistical physics to be examined in real neural tissue. A series of experiments to test hypotheses about stable states and the critical point are proposed to advance this research. The data from these experiments will be used to construct complementary network simulations that should to lead to further testable predictions. Knowledge gained from these experiments is expected to advance the understanding of computational principles used by networks of living neurons and to be useful in designing synthetic devices that exploit these principles. These studies are expected to have broader impact in two areas. First, analysis tools developed for these culture experiments are expected to be applicable to data from whole behaving animals as well. Second, this research will contribute to training interdisciplinary scientists who are expected to be valuable in interpreting the deluge of neural data that will certainly come as recording technology and computer capacity continue to improve.

A central task in understanding how neurons collectively process information is to map how neurons influence each other in local cortical networks. As defined here, local cortical networks will consist of tens to hundreds of neurons. Influence will be defined as how well knowledge of activity in one neuron will allow the activity in another neuron to be predicted. Three methods for measuring influence between neurons will be explored. To assess these methods, they will be used on data from simple, and then realistic, models of cortical networks where the underlying connectivity structure is known. After refinement, the methods will be applied to recordings from hundreds of cortical neurons in small slice cultures of brain tissue. Over 100 cortical neurons at a time will be recorded through the use of an advanced, 512 electrode array. In addition, measures of influence will be applied to data taken from 16 wire electrodes placed in behaving rats. These in vivo recordings will serve as a first step toward linking influence maps in cortical networks to behavior. This research is expected to provide new knowledge that could aid the design of brain-like computing devices. In addition, it could ultimately be used as a tool to identify differences in influence patterns between healthy and pathological brains.

The three methods for measuring influence will include directed information, transfer entropy, and Granger causality. Special care will be taken to identify situations where these measurements may produce false positive connections. These include cases where two neurons are driven by a common source at different delays, and cases where one neuron influences another neuron indirectly through an intercalated neuron. Such false positive connections will be identified and corrected, to the extent possible, by comparing raw pairwise measures of influence with conditional measures of influence. Simulations will also provide an estimate of how often neurons outside the recorded population can contribute to false positive connections. These estimates will be used to place confidence limits on the influence maps extracted from actual data. In neurons where influences converge, synergistic interactions between influences will be measured. The map of influence will serve to identify locations within the network where synergistic transformations of information, or computations, occur.

How do cortical neurons collectively interact to process information? A new hypothesis, supported by recent multielectrode recordings in brain tissue samples and in the intact brain, suggests that local cortical networks self-organize to operate near a critical point. At this point, interactions across all spatial and temporal scales can occur, and avalanches of neural activity are generated at all sizes, resulting in approximately power law distributions. Very generic neural network models suggest that information processing will be optimal at this point. This "criticality hypothesis" has significant implications for computation in cortical circuits, and ultimately, for human health. Despite the potential importance of this hypothesis, it has remained untested in local cortical networks consisting of 100 or more synaptically connected neurons. Previous work has relied on local field potential recordings, which do not reveal the number or location of individual neurons generating signals. Previous work also has relied on relatively few (~60-100) broadly-spaced recording sites, which are insufficient to adequately sample a population of synaptically connected neurons. Here, the PI will record from 512 closely spaced electrodes, allowing us to monitor the spiking behavior of hundreds of neurons, many of which are expected to share synaptic connections. The tests the PI will perform of the criticality hypothesis will allow him, for the first time, to determine if networks of cortical neurons are critical or not.

Scientific education and outreach is another vital component of this proposal. The PI will engage thousands of children from economically disadvantaged areas in the development and use of an interactive exhibit for WonderLab, a popular local science museum for K-middle school children. This exhibit is based on electroencephalogram (EEG) recordings and will display children's brain waves, allowing them to modify these waves by biofeedback. This exhibit will be placed in WonderLab museum, which serves 78,000 visitors each year. WonderLab has been named by Parent's magazine as one of the "Top 25 Science Centers in the United States."

An award is made to Indiana University, Bloomington, to acquire a 512 micro-electrode array instrument (512MEA) for recording and stimulating electrical activity in samples of brain tissue. The 512MEA can be used to measure how groups of several hundred neurons send information back and forth to each other. The ability to map information transfer in networks of this size is expected to be very valuable. Many theories predict that information transfer will change in networks of hundreds of neurons after learning, after exposure to drugs of addiction or toxins, after seizures and after traumatic injury. Other theories predict that brain networks process information in a nearly optimal way, an idea that has remained largely untested. Use of the 512MEA is therefore expected to provide new knowledge relevant to understanding how learning occurs, how drug addiction begins, how poisons affect brain health, how epileptic seizures start, how the brain responds to injury, and how the brain optimizes information processing. Ultimately, this work could benefit society by helping to improve teaching and learning in schools, by helping in the treatment of epilepsy patients, people exposed to harmful substances, war veterans, and by suggesting new ways to design brain-like computers. To maximize the number of students and laboratories using this device, it will be rotated between the Department of Physics, the Department of Brain and Psychological Sciences, and the Medical School.

The research and training opportunities that are opened by the 512MEA center on two topics of investigation: (1) emergent properties, and (2) information transfer. Emergent Properties: Many basic emergent properties of the brain like pattern recognition, associative memory, formation of cell assemblies, neuronal avalanches, synchronized pulses, and collective computations are predicted to arise first in populations of hundreds of interconnected neurons. Although most of these phenomena have been predicted for decades, they have remained largely unexplored for lack of proper instrumentation. The 512MEA would allow all of these topics to be researched in detail, many for the first time. Information transfer: It is almost completely unknown how information transfer differs in networks from naive animals and those that have learned; between networks exposed to neuro-active substances and those that have not; between developing networks and those that have matured. Because the 512MEA can record neural activity at millisecond resolution, it can identify which brain cell became active first in a chain of activity. This ability is crucial, as it will indicate the direction of influence between neurons. Optical methods of recording activity between hundreds of neurons often do not have this capability. The 512MEA can thus permit many of these topics to be researched for the first time.

Networks of cortical neurons are clearly organized into layers and columns, but relatively little is known about how these arrangements affect cortical computations. To approach this issue, a 512 micro-electrode array will be used to stimulate and record activity from hundreds of cortical neurons. With this, the inputs and outputs of a cortical network can be experimentally controlled. A recently-developed framework for understanding neural computation known as "reservoir computing" permits the computational power of neural networks to be quantified based on knowledge of their inputs and outputs. The 512-electrode system allows input stimulation to be localized to different cortical layers or columns. Similarly, outputs can be selected by recording from different layers or columns. Thus, the contributions of layers and columns to computations, and the types of computations they perform, can be measured and compared. The results of this research are expected to increase the understanding of how the cortex attains its remarkable computational power. In addition, the results of this work are expected to inform future designs of brain-like computing circuits. To promote scientific education and outreach, an existing software package called "Simbrain" will be further developed and disseminated. This package will allow students from high school level and above to understand how cortical networks transform inputs into outputs as they perform computations.

Three specific aims will be pursued. First, the measurement of computational capacity must be based on realistic levels of random background stimulation. The high-conductance state is a well-known phenomenon in vivo resulting from constant random synaptic inputs, and is also a common feature in many (particularly reservoir computing) neural circuit models. The 512-electrode array will be used to deliver background stimulation to determine levels that will improve computational performance. Second, layer input and output locations will be studied. Using kernel quality and VC-dimension metrics, the computational power and role of each layer taken individually or as a whole will be assessed. It is possible that some layers more strongly generalize input patterns while others separate them. Thus it will be possible to dissect the computational contribution of each layer. Third, the same metrics will be applied to stimulation to one column which feeds to another. Here the computational power and role of multiple columns will be assessed, and any computational differences between columns directly stimulated by the array and columns stimulated by other columns can be observed.