Project Summary/Abstract
Patterns of neural activity underlie information processing in the brain. Most work to date has focused
on separate stages of computation by looking at separate regions in the brain - one at a time. We
propose that techniques from graph theory can help us better understand how information is processed
by entire populations of neurons. To this end, we use the nematode Caenorhabditis elegans to study
the processing of an ecologically-relevant signal in most of the nervous system at once. Specifically,
we will record activity from all of the neurons in the head of the worm, where most olfactory processing
occurs, while we expose the animal to an innately attractive odor, diacetyl. We will then test how this
representation changes in two behavioral states – after adaptation, at which point the worm no longer
finds diacetyl attractive, and when C. elegans recovers from adaptation, when it again finds diacetyl
attractive. We will do all of this in a transgenic worm which will allow us to identify all neurons by name,
and thus to analyze our results based on the known anatomical connections between neurons. Work
we are preparing to submit for publication has established that one graph-theoretic feature can identify
a stimulus’ valence, i.e. whether or not it is attractive or repellent, and we will determine which neurons
are driving changes in this feature. Finally, we will optogenetically test the predictions from our analyses
to ensure that they are biologically significant. For instance, some non-overlapping subsets of neurons
may represent positive and negative valence, and their activation may induce either forward, or
backward, movements, respectively. If, for example, a neuron provides an important link between
neurons that represent any valence (i.e. it is on the shortest path between these neurons) and the motor
command interneurons, then we might reason that it facilitates the transfer of information, and that
inhibiting it would delay the animal’s odor-seeking behavior. A future extension of this work would
combine graph theory and information theory to understand how efficiently neurons process and
transfer information. Importantly, this field, called network coding, proposes that an efficient way to
transmit information is to allow downstream nodes to decode information that is processed along the
path. During my postdoctoral work, as I gain experience with new theories and a new neural system
that uses spiking neurons, I seek to develop the field of network coding for neuroscience. I am
interested in spiking neurons to ensure my work is applicable to the larger field of neural systems which
employ spiking, not graded, potentials.