Olfactory Coding: Myths, Models and Data

Christiane Linster, Boston University

Wayne Getz, University of California, Berkeley

Frank Grasso, Boston University

This workshop will be held as part of the Workshop Sessions
at the Neural Information Processing meeting
Friday and Saturday, December 4-5 in Breckenridge, CO


What we will talk about ...

Recent data from molecular biology have made realistic the idea of "odor mapping" in the olfactory system, challenging the interpretation of the wealth of previously existing electrophysiological and behavioral data as well as theoretical models of olfactory processing. Briefly, these new data suggest that functionally identical (expressing the same putative receptor(s)) olfactory sensory neurons cells project to the same glomeruli in the olfactory bulb, activating the same postsynaptic neurons. Glomeruli have long been thought of as "functional units" of the olfactory bulb, and these data support that idea. As a consequence, hypotheses for odor coding involving a "one glomerulus - one odor" relationship have become attractive. Experimental approaches involving subgroups of functionally related odorants (for example aliphatic alcohols of varying carbon chain) have reinforced the ideas of receptive fields of olfactory sensory neurons and of mitral cells in the olfactory bulb. Taken by themselves, these types of experimental results push a simplified model of odor processing with many similarities of models of visual processing. However, these hypotheses are challenged by experimental data showing that despite the apparently simple anatomical mapping (one subset of olfactory receptors-one glomerulus), the representations of odors at the level of the olfactory sensory cells is much more complex: (1) more than one functional receptor may be expressed in a given subset of olfactory sensory neurons, (2) a given olfactory sensory neurons responds to a large number of chemical stimuli (up to 30% of tested compounds in most experimental preparations), (3) the number of compounds activating a given receptor neuron depends on the stimulus concentration, (4) both excitatory and inhibitory responses can be observed in olfactory sensory cells and (5) non linear synergistic effects to odor mixtures can be seen as early as at the level of the individual olfactory sensory neuron. Similarly, the data suggests complex non-linear representation of odors at the level of the olfactory bulb: (1) at the level of glomerular input, stimulation with a given odor compound results in activation patterns involving several large areas of the olfactory bulb, the spread of these areas depends highly on concentration and odor mixtures result in activation patterns which are not predictable by the patterns evoked by the components of the mixture, (2) individual mitral cells responds to large numbers of compounds, (3) the so called receptive fields of these mitral cells in response to a subgroup of functionally related odorants are highly dependent on concentration and (4) the response are shaped by two types of inhibitory circuits acting on olfactory bulb mitral cells.

What we want to achieve ...

Despite the large number of theoretical models published in the last ten years, the communication between modelers and experimentalists is sparse in the olfactory community. In this workshop we will bring together a group of experimentalists and a group of theoreticians to discuss the implications of the recent molecular biology data for the interpretation of eletrophysiological and behavioral data pertaining to theories of odor representation and coding. The main focus will be the interpretation of behavioral results with models of odor representation at the level of the olfactory sensory neurons, the olfactory bulb and the olfactory cortex. How are the theories about olfactory bulb processing put forward by modelers related to the experimental knowledge to date? What are the specific hypotheses put forward by each theory and how are they challenged by the data? Can we propose novel interpretations of the experimental data by use of these models? And how can the experimental work in progress be guided by careful theoretical studies?

The practical goal of the discussion around the topics mentioned below is the design of a hypothetical set of experiments, involving electrophysiology, behavior and theoretical modeling that would allow to answer some of the raised questions. We will simulate a group effort of a common research goal that all participants can agree on and can contribute to.

More precisely ...

We propose a two day workshop in which speakers from both experimental and theoretical backgrounds will address a number of functional questions concerning olfactory processing.
The main points we will focus on:

(1) It has been shown recently that olfactory sensory neurons can express more than one putative olfactory receptor. How does that challenge our current models of odor mapping and how can existing data be interpreted in that light?
(2)  What role does non-linear mixture-interaction at the level of olfactory sensory neurons play for odor processing? How can we measure the relative contribution of mixture interaction at the level of the olfactory sensory neurons and at the level of olfactory bulb neurons for the perception of odorants?
(3) What is the nature of the representation of odorants in the olfactory bulb and what role does inhibition play?
(4) Is there plasticity at the level of the olfactory bulb? Can the models of learning in the olfactory bulb proposed by theoreticians be interpreted in the light of the experimental data? What would be he mechanisms and roles of such a plasticity?
(5) If learning exist in the olfactory bulb, what is left for cortex?
(6) How can neuromodulators influence odor representations in theory and in the data?

How we will go about it ...

The two day workshop will have a small number of focussed presentations during each session. Speakers will be asked to address one or several of a set of questions, and to show how their own work addresses these questions in the light of other data and modeling results. Discussions between talks will address how the issues raised fit in to our general understanding of olfaction, how they contradict or concur with other data. At the end of each session, a group of discussant will attempt to agree on a common interpretation of data and theory on the questions addressed during that session.

Who is coming  thus far ...

The preliminary list of speakers, discussants and participants includes

Peter Cariani, Harvard Univ.
Thom Cleland, Tufts Univ.
Stuart Firestein, Columbia Univ.
Wayne Getz, UC Berkely
Frank Grasso, Boston Univ.
David Horn, Tel Aviv
Leslie Kay, Caltech
William Lemon
Michael Leon, UC Irvine
Christiane Linster, Boston Univ.
Jean-Pierre Rospars, INRA, Paris
John Scott, Emory Univ.
Michael Singer, Yale Univ.
Brian Smith, Ohio State Univ.
Mark Stopfer, Caltech
Joel White, Tufts Univ.

If you want to participate ...

Send email to Christiane Linster at cl243@cornell.edu with a short description of your research interest and in what form you would like to participate in our workshop (speaker, discussant, participant, or other).

Practical information ...

Can be obtained from the NIPS webpage