One of the most exciting unsolved problems of biomedical science is how the cellular and network properties of individual neurons, and the information they convey, give rise to the complex behavior of the brain. This fundamental question is examined in the lab by synthesizing state-of-the-art neurobiology with informatics: the science of information underlying both brains and machines.
The Lab's very newest project addresses a major barrier to progress in neuroscience. Contemporary neuroscientists record increasingly larger numbers of signals from nervous systems and correlate them with many behaviors and with disease states. However, this new capacity to acquire signals simultaneously from a hundred or more neurons or networks far outstrips our ability to analyze and to ask questions about these data, severely limiting progress. To address this imbalance, we have just begun to apply new massively-parallel computer technology to proven analytic methods. Such innovative enhanced analyses will enable neuroscience laboratories to derive greater insight from their data, and so advance our understanding of the relation of neural signals and brain function to sensation, perception, decision, and action. Our major impact, aim, and significance are encapsulated in the acronym NEAT: the Neurophysiology Extended Analysis Tool.
The software to address this critical barrier to progress in the field is based on the multiple verified algorithms and proven design of our Spike Train Analysis Toolkit, described below. We are extending the STAToolkit so that it leverages the very new technology of low-cost, drop-in, massively parallel processors. This novel architecture-derived from graphics processing units (GPUs)-is highly compatible with the standards and goals of high-performance scientific computing, yielding supercomputer-level 500 GFLOPS to 1 TFLOPS performance from a single inexpensive card. We will adapt several highly significant neurophysiology measures-including our Toolkit algorithms-to this new architecture, utilizing innovative and transformative programming techniques and extensive testing that are required to achieve transformative impact.
All our recent work described below has been funded by the NIH, including the NIH Blueprint for Neuroscience Research, and NIMH via the Human Brain Project, with past support from NIMH, NINDS, and NSF.
Our most recent completed project developed and implemented parallelized computational algorithms to explore the information content of spike trains and other neuronal signals, towards an understanding of the neural coding underlying visual and somatosensory processing.
The major deliverable of this work, the Spike Train Analysis Toolkit, has have been downloaded by over 1,700 labs and is in active use by neurophysiologists across the globe.
This project brought to bear local and external collaborators, with local resources (our databases and a 64-processor computational array), to explore informational aspects of neural coding and processing. Via both user-specified and project-developed algorithms, we enable analyses to be performed either on-the-fly during dataset submission, or on archived data, permitting post-hoc examination as well as searches for specific patterns of brain activity. A major goal is development of this new field of computational neuroinformatics.
The project coordinates the efforts of brain researchers, computer scientists, and mathematicians at Cornell and beyond. Our many collaborators aid development and testing of access and query methods and viewer tools and provide complementary physiological data from several techniques and preparations.
Our laboratory concentrates as well on a continuing multi-year initiative that develops networked databases of brain neurophysiology to explore two fundamental questions of neuroscience: neuronal identity and coding of neuronal signals. Our cortical neuron database includes somatosensory cortical neurons and characteristic neurophysiological data encapsulating these neurons' responses to specific stimuli.
The data--recordings from neurons in awake behaving brains--include metadata incorporating parameters used by neurophysiologists to describe recording methodology, stimulating paradigms, and electrophysiological responses. To make the database useful to brain neuroscientists, a suite of multiplatform tools supports acquisition, query, and visualization of single and multi-electrode spike train datasets. These tools allow integration of data characterizing responses of cortical neurons to complementary stimuli, synthesizing a unified understanding of brain information processing. With the resulting enhanced utilization of data, experiments can be coordinated among laboratories, conserving valuable and respected species. All data structures and methods defined in this project are designed to be generalizable to many electrophysiological studies in cortical and subcortical structures of the brain.
Our methodology includes development of object-oriented database schemas for neuronal data, as well as the use of Java, permitting databases to be accessible via the Web to any member of the international neuroscience community using any contemporary computer system, including Macintosh, linux or other flavors of UNIX, or MSWindows.
A major prior project initiated development of a central resource for neuroscientists to discover and explore the wide span of web-accessible neurodatabases, computational tool sites, and portals providing neuron and brain information and materials. This Neuroscience Information Framework was developed for the NIH by a multi-institution consortium directed by Weill Cornell's Laboratory of Neuroinformatics. This work is carried on at neuinfo.org and nif.nih.gov.
Believing strongly that informational, computational, or theoretical biology should never be divorced from experimental work, this thrust also continues my laboratory's long-standing interest in neural networks, their neuronal and synaptic components, and their emergent properties. Using techniques I developed and introduced for simultaneous voltage-clamping of multiple interconnected neurons, we will analyze the information carrying and processing capabilities of parallel channels formed by paired Aplysia neurons. These form a testbed, bridging the gap between the single neurons characteristic of invertebrates and the massively parallel columns and modules found in mammalian brains. Related experiments may test aspects of the fire-together, wire-together hypothesis. This work descends as well from studies of interneuronal organization begun 45 years ago in the laboratory of Eric R. Kandel.