Computational Neuroanatomy Minisymposium

Experimental Biology 2000 Meeting
American Association of Anatomists
The Minisymposia Program

Session #: 9002-9
Date: 4/17/2000
Time: 10:15a - 12:45p
Chair: Giorgio Ascoli

Below please find a preliminary program. Click on title to see the abstract, or on the presenting author's name to email comments. Titles, abstract, and lists of authors officially submitted to EB2000 are not necessarily identical to those reported in this web page.


10:15-10:25 Welcome and introduction (Giorgio Ascoli)

10:25-10:40 "The form of motoneurons maximizes the volume of tissue invaded, weighted by connectivity to soma" (Bill Marks).

10:40- 10:55 "The Need for Computational Neuroanatomy: Neuromorphology's Shaping of Neurophysiology" (Jeff Krichmar)

10:55-11:10 "The Modeler's Workspace and neuronal databases" (Mike Hucka)

11:10-11:25 "Modeling dendritic branching patterns, axonal navigation and synaptic competition" (Jaap van Pelt)

11:25-11:40 "Predicting computational properties of neuronal ensembles using a database of reconstructed neurons" (Gwen Jacob)

11:40-11:55 "Architecture of inhibitory and excitatory synaptic connections in the neocortex" (Henry Markram)

11:55-12:10 "The microscopic structural development of the postnatal human cerebral cortex: the CyberChild database" (Rod Shankle)

12:10-12:30 "The computational representation and analysis of neuroanatomical knowledge: limiting the problem of information overload in neuroscience" (Gully Burns and Claus Hilgetag)

12:30-12:45 Panel Discussion

12:45 - ? LUNCH (more info TBA) and INFORMAL DISCUSSION



Progress and perspective in computational neuroanatomy
Giorgio A. Ascoli (Dept. Psychology & Krasnow Institute for Advanced Study, GMU, Fairfax, VA)

Anatomy plays a fundamental role in supporting and shaping nervous system activity, yet to date a complete picture of the details of such a role has escaped the efforts of experimental and theoretical neuroscientists. The tremendous increase in processing power of personal computers has recently allowed the construction of highly sophisticated models of neuronal function and behavior. We implemented complete computer models of dendritic morphology to generate virtual neurons that are anatomically equivalent to their real counterparts. From a restricted and already available experimental database, stochastic and statistical algorithms can create an unlimited number of non-identical virtual neurons within several mammalian morphological classes, storing them in a compact and parsimonious format. When modeled neurons are distributed in 3D and biologically plausible rules governing axonal navigation and connectivity are added to the simulations, entire portions of the nervous system can be “grown” as anatomically realistic neural networks. These computational constructs are useful to determine the influence of local geometry on system neuroanatomy, and to investigate systematically the mutual interactions between anatomical parameters and electrophysiological activity at the network level. A detailed computer model of a “virtual brain” that was truly equivalent to the biological structure, could in principle allow scientists to carry out experiments that could not be performed on real nervous systems because of physical constraints. Such a computational approach to neuroanatomy has a great potential to enhance the intuition of investigators and to aid neuroscience education.

The form of motoneurons maximizes the volume of tissue invaded, weighted by connectivity to soma
William B. Marks, Robert E. Burke (LNLC, NINDS, NIH, Bethesda, MD)

One simple idea to explain the form of neurons, as exemplified by motoneurons (MNs), is to assume that neurons maximize the total current delivered to the soma for a given cell volume (v, fixed as the average for MNs).  If the spatial distribution of synaptic sources is not limiting, we can construct model neurons in which maximum current is proportional to the sum of membrane area elements, dA, weighted by the fraction, c, of somatic current that they deliver: cA=Sum(c dA).  However, models with maximum cA were unrealistically compact (dendrite span 1/8 that of MNs).  We therefore assumed that source density is limiting, requiring dendritic extension. Current to soma in such models is proportional to external volume within some distance, R, of dendritic branches, again weighted by c: cV=Sum(c dV). A search over all possible shapes whose volume = v shows that those with maximal cV branch, and when R = 150 µm, exhibit similar shape parameters to MNs (total spread, branch diameters and lengths, diameter ratios at branch points, and spacing between branches) and have cV less than double that of real MNs.

The Need for Computational Neuroanatomy: Neuromorphology's Shaping of Neurophysiology
Jeffrey L. Krichmar (The Neurosciences Institute, 10640 John Jay Hopkins Drive, San Diego, CA 92121)

Little is known about the influence of morphological variability on the physiological response of neurons. It is assumed that the dendritic variability has an effect on the neuronal response, but it has not been thoroughly investigated. Furthermore, computational neuroscience models take great pains in describing the electrophysiology of neurons while for the most part ignoring the neuronal structure. We investigated the effect of morphological differences, within the hippocampal CA3 pyramidal cell family, on neuronal response. We took morphological measurements from 3-D neuroanatomical data of CA3 pyramidal cells, converted the data into a computational simulator format, distributed channels equally across different cells and tested the simulated neuron's physiological response. Differences in the dendritic morphology of CA3 pyramidal cells had a significant effect on the electrophysiological response. Firing rates of the simulated neurons differed based on their size and shape. Qualitative shifts in response (spiking vs. bursting) were sensitive to the dendritic path length from the tips to the soma and the change in branch diameter as a function of the distance from the soma. These results highlight the importance of accounting for morphology in electrophysiological and simulation studies.

The Modeler's Workspace and Neuronal Databases
Michael Hucka, Jenny Forss, Sara Emardson, David Beeman, James Bower (Division of Biology 216-76, California Institute of Technology)

The central motive of our project is to provide the general neuroscience community access to the information contained in realistic neural models. Our objective is to develop software tools to make the wealth of information already accumulated within the GENESIS simulator about the structural organization of the nervous system more easily accessible and more generally available. By basing our efforts on an existing neural simulation system we are addressing inherent limitations in databases including: a) the accuracy and relevance of the data entered; b) the problem of conflicting data; c) data compression; d) promoting participation in development and use of the database; and e) the connection between the data and its functional significance. We believe that this approach has the potential to provide a much broader group of neuroscientists with access to neurobiological information and a better understanding of the functional organization of the nervous system.

Modeling neuronal geometry and competition in nerve connection development
Jaap van Pelt, Arjen van Ooyen, Harry B.M. Uylings (Graduate School Neurosciences, Amsterdam Netherlands Institute for Brain Research, Meibergdreef 33, 1105 AZ Amsterdam, The Netherlands)

Neuronal branching patterns are complex and show a large degree of variation in their shapes, within and between different cell
types and species. Neuronal structure emerges during development by way of dynamic behavior of growth cones, located at the tip of outgrowing neurites that mediate branching and elongation. The emergence of dendritic complexity is studied on the basis of randomly branching and migrating growth cones. Using mathematical modeling, it is shown that such randomness is sufficient to account for the observed variations in the number, the connectivity pattern and the length of dendritic segments. It is assumed that branching probabilities depend on the position of the growth cones in the tree and decrease with the increasing number of segments in the growing tree [1]. The modal shape of segment length distributions has been described by assigning an initial length at the time of branching to the newly formed segments. Different mean elongation rates are predicted for two distinct phases of rat cortical pyramidal dendritic development [2]. Excellent agreement has been obtained in the dendritic shape parameter distributions for a variety of cell types including rat cortical pyramidal cell basal dendrites, cat superior colliculus neurons, and Guinea pig Purkinje cell dendritic trees. The mathematical modeling of dendritic growth on the basis of random behavior of growth cones has brought many dendritic shape parameters and their variations into a coherent framework. The development of connections between neurons and their targets involves competition among innervating axons for target-derived neurotrophins. Although the notion of competition is widely used within neurobiology, there is little understanding of the nature of the competitive process and the underlying mechanisms. We present a new theoretical model to analyse competition in the development of nerve connections [3]. According to the model, the precise manner in which neurotrophins regulate the growth of axons (in particular the increase in the axon's total amount of neurotrophin receptor), determines what patterns of target innervation can develop. The regulation of neurotrophin receptors is also involved in the degeneration and regeneration of connections. Competition in our model can be influenced by factors dependent on and independent of neuronal electrical activity. Our results point to the need to measure directly the specific form of the regulation by neurotrophins of their receptors. Diffusible chemoattractants and chemorepellants, as well as contact attraction and repulsion, have been implicated in the establishment of connections between neurons and their targets. Here we study how such diffusible and contact signals can be involved in the whole sequence of events from bundling of axons, guidance of axon bundles towards their targets, to debundling and the final innervation of individual targets. By means of computer simulations, we investigate the strengths and weaknesses of a number of particular mechanisms that have been proposed for these processes [4]. (References: [1] Van Pelt, J., Dityatev, A.E. and H.B.M. Uylings. Natural variability in the number of dendritic segments: Model-based inferences about branching during neurite outgrowth. J. Comp. Neurol. 387:325-340, 1997. [2] Van Pelt, J. and H.B.M. Uylings. Natural variability in the geometry of dendritic branching patterns. In: Modeling in the Neurosciences: From Ionic Channels to Neural Networks, R.R. Poznanski (Ed.), Harwood Academic Publishers, Amsterdam, p.79-108, 1999. [3] Van Ooyen, A. and D. J. Willshaw. Competition for neurotrophic factor in the development of nerve connections. Proc. R. Soc. Lond. B (1999) 266: 883-892. [4] Hentschel, H. G. E. and A. van Ooyen, A. Models of axon guidance and bundling during development. Proc. R. Soc. Lond. B. in press)

Predicting computational properties of neuronal ensembles using a database of reconstructed neurons
Gwen A. Jacobs (Center for Computational Biology, Montana State University, Bozeman, MT 59717)

Neurobiologists have known for decades that to understand the computational properties of a neural system, it is necessary to understand the relationships between the physiological properties of individual neurons and their anatomical structures. We have addressed this problem by developing an approach for analyzing the relationships between structure, function and computation within a network of neurons. We have developed a suite of tools, called NeuroSys, ( that allow the investigator to reconstruct the anatomical features of many neurons and store them in a database that preserves their correct spatial relationships in the nervous system.  This ensemble reconstruction can then be used as a precise anatomical template on which to predict connectivity patterns and image the functional properties of the network. The visual format of the database is a probabilistic atlas, which preserves the spatial relationships between all objects within the nervous system. The database can be queried for information regarding structural, functional and relational attributes of the objects, and be used to predict functional properties of the neural system. The goal of the NeuroSys system is to provide an interactive environment to build realistic models of neural networks that incorporate both the precise anatomical relationships between neurons as well as their physiological properties.  These models can then be used as predictive tools to understand the functional organization of neuronal networks. The NeuroSys suite of programs was developed using a model sensory system in an insect, the cricket cercal sensory system (Troyer, et al 1994). Neurons were reconstructed in 3 dimensions and then scaled and aligned to a common coordinate system (Jacobs and Nevin, 1991).  The file format of the 3D reconstructions is a vector-based branched tree structure.  Each segment of the neuron is represented as a cylinder, with a set of X, Y, and Z coordinates and a diameter.  This format is extremely useful for a variety of quantitative analyses, including compartmental modeling and quantitative morphometrics. NeuroSys has a variety of visualization tools for examining the structural relationships between neurons. The anatomical relationships between neurons can be quantified by calculating a statistical estimate of the membrane surface area of the neurons (Jacobs and Theunissen, 1996). These density distributions have been used to quantify the anatomical overlap between different neurons in the database and to predict the probability of synaptic connectivity between pre and post synaptic neurons. These tools can also be used to predict and analyze the computational properties of the network. Any functional attribute stored in the database can be assigned to individual neurons in the ensemble. The anatomical mapping of a variety of functional attributes can be studied by viewing the entire ensemble of neurons. We have studied the mapping of multiple functional parameters within the cercal system using this technique ( Paydar et al 1999).  This technique has been used to predict 1) the connectivity relationships between primary sensory afferents and sensory interneurons and 2) the steady state response patterns of the ensemble of neurons to sensory stimuli.  This technique has been extended to incorporate the dynamic patterns of activity of the sensory afferents to predict the spatial and temporal aspects of the response patterns.   To predict the firing patterns of primary sensory afferents, we used a forward model derived from physiological experiments.  This model can be used to predict the firing pattern of a sensory afferent in response to any arbitrary stimulus.  The firing pattern of each neuron in the ensemble was combined to animate the spatial patterns of activity produced by the ensemble.  In future work these ensemble patterns of activity will be used as synaptic inputs to compartmental models of primary sensory interneurons. The ultimate goal of our work is to develop tools that will enable neuroscientists to understand the functional organization of networks of neurons involved in a variety of computational tasks.  These tools will elevate databases from the role of a sophisticated lab notebook to an essential and enabling tool for the formulation and testing of hypotheses related to the mechanistic basis of neural computation, plasticity and development. (References: [1] Jacobs,G.A. and R.Nevin (1991) Anatomical relationships between sensory afferent arborizations in the cricket cercal system.  Anatomical Record 231:563-572. [2] Jacobs, G.A. and F.E. Theunissen (1996)  Functional organization of a neural map in the cricket cercal sensory system, J. Neuroscience 16: 769-784 [3]
Paydar, S. Doan, C. and G.A. Jacobs (1999) Neural Mapping of Direction and Frequency in the Cricket Cercal Sensory System.  J. Neuroscience 19: 1771-1781 [4] Troyer, T.W., Levin, J.E and G.A. Jacobs (1994) Construction and analysis of a data base representing a neural map.  Microscopy Research and Techniques  29:329-343


Estimating cortical connectivity from statistical properties of the microscopic features of the developing human cerebral cortex
Junko Hara1,4,6, James H. Fallon3, Ryuta Fukuda4, A. K. Romney5, and William R. Shankle2 (1Division of Biology, 216-76, California Institute of Techonology, 1200 E. California Blvd., Pasadena, CA 91125; 2Dept. of Cognitive Science, University of California at Irvine, 92697-5100; 3Dept. of Anatomy and Neurobiology, University of California at Irvine, 92697; 4Bioinformatics Lab., Keio University, 5322 Endo, Fujisawa, Kanagawa, 252-0816 Japan; 5School of Social Science, University of California at Irvine, 92697-5100; 6Dept. of Information and Computer Science, University of California at Irvine, 92697)

The relatively simple and modular design of neurons in mammalian cerebral cortex limits the complexity of an individual neuron’s information processing ability.  Rather, complex functions such as language and cognition in humans are accomplished through networks created by their synaptic connections. Connectivity data are therefore very important to understanding higher mammalian brain function. Such data from non-human mammalian species provide many insights regarding the ability of biological neural networks to perform complex tasks (Young et al., 1995, Scannell et al., 1999, Hilgetag, et al., 1996).  However, because of the additional structures and their interconnections that have evolved in the human brain, experimental data on nonhuman primates are insufficient to understand cortical functions unique to or significantly advanced in humans. Because of the paucity of neuroanatomical tracing studies of human cerebral cortex, studies of human cortical connectivity are largely inferred from non-human primate tracing data and from human functional imaging data.  Microscopic, neuroanatomical methods that could be applied on a large scale to indirectly estimate cortical connectivity could significantly accelerate progress in this area. The present study introduces such a method to show the potential of using quantitative microscopic, neuroanatomic data for at least 32 different cortical areas over postnatal human development from birth to 72 months to indirectly estimate cortical connectivity. The results agree well with those of human functional imaging studies.  Comparisons with these functional imaging studies and with non-human cortical connectivity data will be presented.

The computational representation and analysis of neuroanatomical knowledge: limiting the problem of information overload in neuroscience.
Gully A.P.C. Burns  &  Claus H. Hilgetag (Univ. of So. CA and Boston University)

Practitioners of the discipline of Neuroanatomy suffer greatly from information overload, due to the extent, the complexity, as well as the complicated taxonomy of their subject. We describe the basic concepts of a computational approach to managing and synthesizing descriptions of neuroanatomical connections into a coherent data model. This approach uses mathematical analyses of neural connectivity to reveal fundamental principles of brain organization. The task of building a representation of neuroanatomical circuitry may be described as an exercise in knowledge management, i.e. the management of organized facts within the conceptual framework of human interpretation and understanding. This process involves processes of prioritization and intuitive deduction, which currently cannot be performed artificially, but which may be represented in a computational knowledge base management system. Our system (called 'NeuroScholar') can be used to construct explicit representations of neuroanatomical knowledge. This knowledge can then be analyzed formally with data-mining techniques to investigate principles of neuroanatomical organization. The system is flexible enough to accommodate representations of more generalized neuroscientific knowledge, so that functional data may be included in future analyses. We present results describing the organization of neural circuitry in the rat and Macaque monkey brain, obtained with a new computational analysis approach based on evolutionary optimization. Our analyses revealed that sensory cortical systems in mammalian brains possess a strict yet flexible hierarchical architecture, and that functional and anatomical connectivity networks in the cortex form connectivity clusters, which are specific for sensory and functional modalities. Given the wide range of - partly very surprising - results that can be obtained through the computational representation and analysis of neuroanatomical knowledge, we suggest that these computational approaches may lead to revolutionary theoretical developments within neuroscientific research.