This subproject is one of many research subprojects utilizing theresources provided by a Center grant funded by NIH/NCRR. The subproject andinvestigator (PI) may have received primary funding from another NIH source,and thus could be represented in other CRISP entries. The institution listed isfor the Center, which is not necessarily the institution for the investigator.Neurons use electrical signals to communicate with each other. The time-varying pattern of these neuronal signals (firing dynamics) is critically important to information coding. Neuronal firing dynamics are largely determined by the excitability of the neuronal membrane, driven by active ion channels, and the shape (morphology) of its dendrites [1]. The relative importance and interactions between active and morphologic properties remain poorly understood, particularly since these properties, and even the firing dynamics they determine, may vary during development, learning, aging and disease [2-4]. In computational models, active and morphologic features interact nonlinearly, represented by many parameters with different magnitudes and even different units of measurement. I am a member of the research faculty at Mount Sinai School of Medicine, working in the laboratory of Prof Susan Wearne, Associate Professor of Neuroscience. Our research program is twofold: to design computational techniques to investigate scientific questions methodically [5-10], and to use these techniques to study how the organization of intrinsic neuronal properties influences function [10-12]. This particular project is supported by an NIH grant (DC05669) to Prof Wearne. Working memory, which maintains a brief mental representation of a recent event necessary for future task performance, underlies many of the brains complex functions. The Wearne laboratory studies properties of networks and individual neurons that participate in working memory tasks [6,7,10-13]. For example, persistent neural activity, lasting long after an input has ended, is a hallmark of working memory observed throughout the brain [14]. In the vestibular (balance) system, persistent activity is needed to maintain precisely tuned eye velocity in response to a given angular head velocity. In the goldfish, neurons from the hindbrain nucleus Area II exhibit, and likely contribute to, this eye velocity storage [15-17]. Area II neurons display many interesting features in vivo and in vitro [13,17]. Recent models have hypothesized how morphology might contribute to persistent activity [18-20], but none have investigated the contributions of realistic 3D morphology. As morphology varies widely in Area II neurons [16], we have worked to evaluate how morphology influences the precisely tuned firing patterns that underlie persistent activity in Area II. Using the computational methods described below, we have demonstrated that morphology can contribute significantly to neuronal firing properties [10-12,21,22]. In Area II dendrites the tapering, flare and diameter contribute significantly to a range of firing dynamics, as do the persistent sodium, calcium-dependent potassium, and A-type potassium membrane conductances [10,21,22]. We are now extending these results based on simple morphologic models to our realistic morphologic data and physiology collected in vivo and in vitro. The computational resources needed to conduct this research have increased dramatically in our move to more realistic models. To meet this increased computing demand, I would like to obtain a Development Allocation account on one of the supercomputers available through the TeraGrid. Computational Methods Evaluating the contributions of morphology and active ion channels to biologically relevant behaviors requires that all parameters (on the order of 10-100 depending on the system) be assigned values consistent with experimental data. Optimization eases this task by automatically searching a multidimensional parameter space to identify parameter sets that minimize a fitness function representing salient differences between simulated and experimental (target) data. While the results are less subjective and identified more quickly than with hand tuning, automated parameter optimization often requires weeks of computing time to identify suitable parameter sets. My research requires the fitting of model parameters to the complex morphology of three different Area II neurons, as well as several reduced models. To conduct highly efficient searches, I have already designed a fitness function that matches the shape of neuronal firing patterns observed in many neurons [9,10]. At the same time, I have also implemented simulated annealing with recentering, a constrained parameter search algorithm ensuring that identified parameters are biologically plausible [9,23]. After fitting our models to the morphologic and physiologic data of Area II neurons, we will explore how the different model parameters affect model output. To do this, I will use a novel application of sensitivity analysis that Prof Wearne and I have developed [10]. Importantly, our method predicts which parameters to change, and by how much, to compensate for changes in another parameter to restore normal function. These predictions go beyond our Area II models, and may prove important to neuronal aging, disease and trauma research. Exploring these complex models requires significant computation time. To evaluate the sensitivity of model output to each of its many parameters, each parameter must be perturbed independently while others are held constant and a new simulation performed. Needing to explore the sensitivity of several model outputs to 10-20 model parameters, at thousands of local points throughout the parameter space, this study requires considerable amounts of computation time. Computing Requirements This research is already underway on a local resource, a 160-processor G5 Xserve cluster running Mac OS X. For timely completion of this study, I need access to multiple processors for extended periods of time, but will not require access to shared memory machines. So far 500 GB RAM has been sufficient for these simulations; this is not expected to change significantly. My disk space requirement should not exceed 1/3 TB. Access to tape storage is not necessary. I ask for a total of 30,000 SUs, to conduct parameter optimizations and sensitivity analyses of models for each of the three Area II morphology obtained experimentally and for analysis of related reduced models. I will need no more than 120 processors at any one time, and will often use less. No communication is needed between the different processors. The fitness function and simulated annealing algorithm have both been implemented in NEURON [24,25], one of the standard modeling packages within Computational Neuroscience. At present, the simulated annealing code runs several serial jobs simultaneously. In the future, we also look to take advantage of Parallel NEURONs Bulletin Board capacity of running multiple jobs on slave processors, controlled by a single master node, for improved searches of the parameter space. An account on nearly any of the TeraGrid resources would fill these needs. Any of the supercomputing clusters should work well for my needs. NEURON is already running on resources at the San Diego and Pittsburgh Computing Centers (DataStar at SDSC, and BigBen at PSC). If sufficient computing time were available on one of those machines, that might be simplest. Otherwise Michael Hines, one of NEURONs creators, has already offered his assistance in compiling NEURON on any machine to which I am granted access. References 1. Mainen ZF, Sejnowski TJ (1996) Influence of dendritic structure on firing patterns in model neocortical neurons. Nature 382: 363-366. 2. Duan H, Wearne SL, Rocher AB, Macedo A, Morrison JH, et al. (2003) Age-related dendritic and spine changes in corticocortically projecting neurons in macaque monkeys. Cereb Cortex 13: 950-961. 3. Bucher D, Prinz AA, Marder E (2005) Animal-to-animal variability in motor pattern prediction in adults and during growth. J Neurosci 25: 1611-1619. 4. Chang YM, Rosene DL, Killiany RJ, Mangiamele LA, Luebke JI (2005) Increased action potential firing rates of layer 2/3 pyramidal cells in the prefrontal cortex are significantly related to cognitive performance in aged monkeys. Cereb Cortex 15: 409-418. 5. Weaver CM, Hof PR, Wearne SL, Lindquist WB (2004) Automated algorithms for multiscale morphometry of neuronal dendrites. Neural Comput 16: 1353-1383. 6. Wearne SL, Rodriguez A, Ehlenberger DB, Rocher AB, Henderson SC, et al. (2005) New techniques for imaging, digitization and analysis of three-dimensional neural morphology on multiple scales. Neuroscience 136: 661-680. 7. Rodriguez A, Ehlenberger DB, Hof PR, Wearne SL (2006) Rayburst sampling, an algorithm for automated three-dimensional shape analysis from laser scanning microscopy images. Nature Protocols 1: 2152-2161. 8. Rothnie P, Kabaso D, Hof PR, Henry BI, Wearne SL (2006) Functionally relevant measures of spatial complexity in neuronal dendritic arbors. J Theor Biol 238: 506-526. 9. Weaver CM, Wearne SL (2006) The role of action potential shape and parameter constraints in optimization of compartment models. Neurocomputing 69: 1053-1057. 10. Weaver CM, Wearne SL (2007) Neuronal firing sensitivity to morphologic and active membrane parameters. PLoS Comput Biol submitted. 11. Kabaso D, Nilson J, Luebke JI, Hof PR, Wearne SL (2006) Electrotonic analysis of morphologic contributions to increased excitability with aging in neurons of the prefrontal cortex of monkeys. 2006 Neuroscience Meeting Planner. Atlanta, GA: Society for Neuroscience. 12. Coskren PJ, Hof PR, Wearne SL (2007) Age-related neuromorphological distortion affects stability and robustness in a simulated test of spatial working memory. BMC Neuroscience 8: P169. 13. Wearne SL, Gamkrelidze G, Weaver CM, Baker R (2006) Distinct modes of spike generation recorded from Area II neurons in goldfish hindbrain slices. 2006 Neuroscience Meeting Planner. Atlanta, GA: Society for Neuroscience. 14. Major G, Tank D (2004) Persistent neural activity: prevalence and mechanisms. Curr Opin Neurobiol 14: 675-684. 15. Pastor AM, de la Cruz RR, Baker R (1994) Eye position and eye velocity integrators reside in separate brainstem nuclei. Proc Natl Acad Sci 91: 807-811. 16. Straka H, Beck JC, Pastor AM, Baker R (2006) Morphology and physiology of the cerebellar vestibulolateral lobe pathways linked to oculomotor function in the goldfish. J Neurophysiol 96: 1963-1980. 17. Beck JC, Rothnie P, Straka H, Wearne SL, Baker R (2006) Precerebellar hindbrain neurons encoding eye velocity during vestibular and optokinetic behavior in the goldfish. J Neurophysiol 96: 1370-1382. 18. Koulakov AA, Raghavachari S, Kepecs A, Lisman JE (2002) Model for a robust neural integrator. Nat Neurosci 5: 77-82. 19. Goldman MS, Levine JH, Major G, Tank DW, Seung HS (2003) Robust persistent neural activity in a model integrator with multiple hysteretic dendrites per neuron. Cereb Cortex 13: 1185-1195. 20. Lowenstein Y, Sompolinsky H (2003) Temporal integration by calcium dynamics in a model neuron. Nat Neurosci 6: 961-967. 21. Weaver CM, Gamkrelidze G, Baker R, Wearne SL (2005) Intrinsic dendritic properties contribute to firing regularity in model neurons of the velocity storage integrator. 2005 Abstract Viewer/Itinerary Planner. Washington, D.C.: Society for Neuroscience. 22. Weaver CM, Gamkrelidze G, Baker R, Wearne SL (2006) Sensitivity of firing dynamics to intrinsic dendritic properties in a model of neurons necessary for eye velocity neural integration. 2006 Neuroscience Meeting Planner. Atlanta, GA: Society for Neuroscience. 23. Cardoso MF, Salcedo RL, de Azevedo SF (1996) The simplex-simulated annealing approach to continuous non-linear optimization. Computers Chem Engng 20: 1065-1080. 24. Carnevale NT, Hines ML (2006) The NEURON Book. Cambridge, UK: Cambridge University Press. 25. Weaver CM (2007) AP shape and parameter constraints in optimization of compartment models. ModelDB. http://senselab.med.yale.edu/senselab/modeldb/ShowModel.asp?model=87473
Showing the most recent 10 out of 292 publications
