Motivations for NemaSys


Traversing scales of organization

An objective of computational neuroscience is to better understand how biological neural networks process information and control behavior.  This is challenging in part because understanding nervous system function requires thinking at many spatial scales of organization.  Researchers often work to relate variables and parameters defined at neighboring scales, with the hope of eventually extending that understanding across many scales.  The complexity of most organisms makes this task prohibitively difficult, but the nematode Caenorhabditis elegans is a simple organism for which a detailed computer representation of the entire body and entire nervous system is an attainable goal.  In this animal it is attractive to seek to relate genetics to behavior, but especially via cellular and network dynamics.


A uniquely interesting organism

The nematode C. elegans, a tiny soil-dwelling worm, is arguably the hydrogen atom of systems neuroscience.  The adult hermaphrodite is only 1 mm long and 80 µm in diameter (Figure 1).  It is comprised of exactly 959 cells, including 302 neurons and 95 muscle cells (Wood, 1988).  An exhaustive study of its neuroanatomy by White et al. (1986) provides an unusually complete characterization of the morphology of every neuron, and the location of essentially every electrical and chemical synapse in the nervous system.  This is the only organism for which we have such detailed information, making it extremely attractive for neurobiological studies.


Figure 1.  The adult nematode C. elegans, shown with its nose at the left.


Despite its small size, C. elegans has a rich behavioral repertoire, making it an excellent model organism for studying the neural control of behavior.  In fact, C. elegans has long been used in genetics research, and an enormous number of mutants have been identified through behavioral variations (Bargmann, 1993).  The Worm Genome Project (, in part a warm-up for the Human Genome Project, was completed in 1999.  Due also to genetic methods, it is possible to mail order worms which have specified neurons glowing with green fluorescent protein (GFP).  This makes laser ablation of identified neurons possible, helping to determine the role of individual neurons in a given behavior.  Laser ablation has been applied to nematode tap reflex by Wicks et al. (1996), to thermotaxis by Mori and Ohshima (1995), and to chemotaxis by Bargmann and Horvitz (1991).  Despite their tiny size, it has even become possible to make whole-cell patch-clamp recordings from identified neurons in C. elegans (Goodman et al., 1998) to determine passive membrane and active ion channel properties. 

A proper integration of these anatomical and physiological data would allow researchers to relate genetics, anatomy, physiology and behavior in unparalleled detail.  This has been recognized and commented on by C. elegans researchers for some time (Wood, 1988), but not yet brought to fruition.  Genetics and behavior have been studied in great detail, but mostly by ignoring the intermediate cellular and network-level dynamical determinants of behavior. We believe the effort necessitates a multi-scale computer simulation environment for C. elegans.


Prior modeling work in C. elegans

Several groups have begun working toward realistic computer modeling of C. elegans.  Neibur and Erdos (1991, 1993) constructed an anatomically detailed model of the nematode body, and showed how Newtonian force equations generate forward and backward movement of a simulated worm placed on a slightly wet agar surface.  The model includes the external cuticle, internal fluid pressure, and body wall muscles controlled by motor neurons.  It is obviously desirable to base all neural network simulations on this realistic body model.  Yet these equations have not been used in any subsequent work to our knowledge, partly because they require some effort to implement, but also because the authors reported at the time that their numerical solutions were very computationally demanding.  To circumvent this, Ferree et al. (1997) derived a simplified body model, based only on the location and orientation of the tip of the nose and the degree of bend in the neck, which determines the rate of turning of a forward-moving worm as a function of relative muscle lengths.  But with the increased speed of modern computers, the exact body equations are no longer computationally prohibitive.  A one-time effort would make the body model readily available for integration with network studies.

Others have also circumvented the body model, and focused on more complex behaviors.  Wicks et al. (1996) developed a network model of the tap-reflex response, based upon neurons inferred from laser ablation experiments.  They put forth a key theoretical argument, that tonic neurotransmitter release, nearly ubiquitous in nematode nervous systems, implies a stable fixed point of the network in the absence of sensory input.  Ferree and Lockery (1999) developed a linearized model in the vicinity of this fixed point, and optimized it directly to control chemotaxis.  The notion of stability in the nematode nervous system needs to be reconsidered in the context of the nearly constant sinusoidal movement, however, and such a study would benefit from a more accurate body model. 


Neural simulators for computational neuroscience

Computational neuroscience draws researchers from a diverse range of scientific backgrounds.  The field has benefited greatly from neural simulators, which allow scientists from many backgrounds to rapidly build biologically realistic neural models, without requiring them to first master and implement a great deal of mathematics, numerical algorithms and computer programming.  Two primary examples are NEURON (Hines, 1993) and GENESIS (Bower and Beeman, 1998).  These simulators are not purported to “solve the brain problem,” but only to provide a flexible programming and visualization environment, with tools written at a much higher level than C or Matlab, and adapted to the specific problem of neural simulation.


© 2001 Thomas Charles Ferree

Computational Science Institute, University of Oregon