February 1, 2008 | David F. Coppedge

Nose Code Rockets Smell Discrimination

You have a code in your nose.  Scientists working on fruit fly olfactory systems have found that a mapping mechanism between components maximizes the fly’s ability to discriminate smells.  The coding system provides a non-linear response that appears finely tuned to maximize the information content of odor inputs.
    The components of this system (antennal lobe, olfactory bulb, glomeruli, projection neurons, Kenyon cells) were described in more detail in our 06/27/2005 entry.  Why is the input information passed through several stages that combine, sort and distribute their respective inputs?  It turns out that a kind of computational algorithm is being performed on the input that translates a highly-variable chemical code into a discrete neural code.  This is analogous to analog-to-digital conversion.  In the process, one part can average away uncorrelated variability, but amplify correlated signals.  The outputs are amplified in a non-linear fashion, providing the maximum discrimination between input smells, some of which might be very similar.  The system also allows the teasing out of information from very weak inputs.
    Baranidharan Ramana and Mark Stopfer, writing in Current Biology,1 commented on a recent study by Bhandawat et al. who “used the relatively simple olfactory system of the fruitfly Drosophila to show how noisy, variable peripheral signals are transformed by early neural circuits into consistent, efficient and distinguishable odor representations.  In their review, they described how the coding is finely balanced:

…Bhandawat et al. found that the responses of projection neurons were highly correlated with each other, as were responses within groups of ORNs [olfactory receptor neurons].  The existence of inhibitory and excitatory local neurons in the antennal lobe suggests that both competitive and associative interactions are possible.  Purely inhibitory interactions between projection neurons would tend to decorrelate their responses; in an extreme case – a fully connected network – such interactions would lead to a ‘winner-take-all’ competition, resulting in a coding capacity greatly reduced to the number of available output channels.  Purely excitatory interactions, on the other hand, would decrease the independence of channels to lower than what is available in their inputs.  Thus, the results of Bhandawat et al. suggest that the network connectivity of the antennal lobe is delicately balanced to optimize its coding capacity.

By optimizing its coding capacity, they mean that the system is sensitive, consistent, efficient, and reliable – despite the “chaotic structures of odor plumes” that are the inputs.  They said that the system also includes a “‘high-pass’ filtering function [that] may allow flies to alter behaviors rapidly when stimulated by odors.”  In addition, the time variability of the input signals gives the system additional real-time information.
    What’s the usefulness of studying a fruit fly antenna?  In short, the elegant solution seen here can inspire human engineers faced with similar information-processing problems:

The work by Bhandawat et al. provides insights into the logic behind olfactory circuit design.  It will be interesting to analyze how these results generalize to a larger set of odorants, and to other species.  It will also be interesting to see whether these peripheral circuits play a role in insulating the neural signal-processing engine from the constant changes in the population of ORNs that occur throughout the lifetime of the animal.  These fundamental olfactory processing principles are not only important for understanding how the brain interprets odor signals, but are also necessary for engineering solutions inspired by biological computations for addressing high dimensional and non-linear problems.

The authors noted that the coding system found in the fruit fly is similar to that used by many animals.  This prompted an evolutionary speculation: “This peripheral reorganization scheme is remarkably similar across species and phyla, suggesting a design optimized over evolutionary time to solve a common information processing problem.


1.  Baranidharan Ramana and Mark Stopfer, “Olfactory Coding: Non-Linear Amplification Separates Smells,” Current Biology, Volume 18, Issue 1, 8 January 2008, pages R29-R32; doi:10.1016/j.cub.2007.10.063.

Scratch that last lame, useless, dull, insipid, witless remark about evolutionary time optimizing design.  Can anyone conceive of how many lucky mutations would be required for this multi-part system to evolve?  The inputs are correlated and decorrelated, converged and distributed, filtered and amplified through a remarkable series of stages that results in a non-linear fashion to gain the maximum amount of information from the input.  If any one of these stages failed, the whole system wouldn’t work.  This is more elaborate than the famous Honda Accord commercial in which a rolling lug nut as input moved a half-ton automobile as output by passing through a finely-tuned chain reaction of alternating strong and weak energetic events.
    Don’t accept balderdash about miracles in evolutionary time brought about by random, mindless mutations that yield optimized designs by accident.  This story was about intelligent design from start to finish.  It should make us stand in awe of creation while we employ our creativity to imitate and apply the engineering for the improvement of human life.  That is the fragrance of science.

(Visited 12 times, 1 visits today)

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.