Keith Godfrey

Research Interests

Spatial representation in the hippocampus and entorhinal cortex

Keith’s interest in hippocampal studies relates to spatial representation, particularly how inputs to the hippocampus generate place responses, and how these place responses feed back to influence the input to the hippocampus. He is also interested in remapping in hippocampus and entorhinal cortex, and what factors (internal and external) influence the formation of maps, the transition between them, and if maps extend beyond a 2-dimensional reference frame. To explore these topics he has began recording from both hippocampus and entorhinal cortex, and hopes to soon begin simultaneous entorhinal cortex/CA3 recordings. In addition to this, he is working on a theoretical model of grid-cell/place-cell interactions that describes how the same mechanisms can represent both novel and familiar locations, and also how representation of novel areas in a familiar environment fits into this framework. This model will also be used to consider the phenomena of pre-play and replay, where hippocampal place responses are activated before and after visiting a location.

Neural development

The complexity observed in the brain is a direct result of the connectivity established during development. This development is largely the result of axon growth, with axons guiding themselves through the neuropile to their target areas, and axon-dendrite interactions generating and refining patterns of synaptic connectivity. Significantly, these activities occur as a result of local interactions, including those leading to such simple decisions of whether a particular synapse should retract or stabilize. Keith's interest is in exploring the functional behaviors that guide axon and synapse development at the local level which lead to the stereotypical patterns of development observed at the network and system levels.

To do this, he has used a modeling approach and focused on visual system development. Noting that neural development is a complex processes, being an emergent property of many contributing cellular and sub-cellular phenomena, he has produced models that represent the many behaviors believed causative to neural development, including axon growth and retraction, spiking neurons, synapse growth and retraction, molecular guidance and activity-dependent trophic feedback, and have show how these behaviors robustly reproduce observed patterns of organization observed at the network level (retinocollicular pathway, wild-type and mutant development). These models were robust in that it was possible to change the way the neural behaviors were implemented while still producing qualitatively similar results. This robustness led to other conclusions about computational models, as if the same general behaviors can be implemented in many ways and yet produce the same results, no individual model can do more than suggest insight into the phenomena addressed and that models are poorly suited for providing information about the mechanistic implementations underlying the phenomena.

Neural engineering

It is Keith’s perspective that computational modeling is more of an engineering than scientific discipline and that traditional computational modeling approaches are inadequate to offer substantive theoretical insight into neural behavior and development. One of the reasons for this conclusion is that most models are exaggerated simplifications of a highly complex system and often with discordant representations of detail, whether that be the wide variations of biological detail represented in a model or the quantitative matching of specific experimental data in a highly abstract model. As an engineering tool, however, computational and mathematical models can have a significant practical impact -- consider for example the development and application of the back-propagation algorithm.

Keith’s goal is to use computational techniques to design neural systems capable of sensing and interacting with their environment, with direct applications in autonomous robotic control. By basing such a system on known biological connectivity and data, we can leverage from nature’s development of dynamic and highly effective control systems. Moreover, producing a fully functional neural-based system will produce stronger theoretical insight into neural behavior and development than is likely from traditional computational models. An autonomous neural system will demonstrate many needs of a functioning network and show how neural phenomena, for example EEG oscillations and the phasing/dephasing of spiking activity, can play important and even critical roles in neural behavior and function. These requirements can provide insight back into the biological system and help guide and/or influence experimental research.

Degrees

  • PhD - University of British Columbia, Graduate Program in Neurosciences.
  • BS - University of Alaska, Department of Mathematical Sciences.

Selected Publications

  1. Godfrey KB, Eglen SJ and Swindale NV (2009) A multi-component model of the developing retinocollicular pathway incorporating axonal and synaptic growth. PLoS Comput Biol 5:e1000600.
  2. Godfrey KB and Eglen SJ (2009) Theoretical models of spontaneous activity generation and propagation in the developing retina. Mol Biosyst 5(12):1527-35.
  3. Godfrey KB and Swindale NV (2007) Retinal wave behavior through activity-dependent refractory periods. PLoS Comput Biol 3:e245.
  4. Godfrey KB (in preparation) Functional modeling of the retinocollicular pathway: retinotopy, segregation and ephrin-A/EphA mutants.
  5. Godfrey KB (in preparation) Considering the strengths and limitations of computational neural models.