PRINCIPAL COMPONENT ANALYSIS OF BEHAVIORAL CHOICE IN THE LEECH CNS.  
K.L.Briggman1*; H.Abarbanel2; W.B.Kristan3  
1. Computational NeuroBiol., 2. Physics, 3. Biol., 
UC-San Diego, La Jolla, CA, USA 
 
We want to find the neuronal basis for making behavioral choices. 
To do this, we are using the leech CNS. We stimulate a nerve in the 
isolated leech nervous system with a fixed stimulus to elicit a 
behavioral choice: sometimes swimming occurs and other times crawling.
We use a FRET-based voltage sensitive dye to record from approximately
100 neurons in a midbody ganglion immediately preceding and following
the stimulus. We use Principal Component Analysis (PCA) to decompose
and reduce the dimensionality of these large data sets. We are using
PCA to answer two questions: 1) Which cells are coactive during
behavioral choices? 2) When is the behavioral choice made? Our data
sets are large 2D matrices of cell number versus time. When using
time as the dimension, we find that 3 principal components account
for most of the variance in our data, which allows us to represent
each cell in a 3-dimensional space. Each cell in this 3D space is
assigned a unique color, such that each color represents a temporal
trajectory. These colors are then projected onto the spatial map of
the ganglion. This representation captures both spatial and temporal
information, which aids in finding coactive cells both within and
between individual trials. When using cell number as the dimension,
we reduce the dimensionality down to 3 activity profiles, where each
profile is a snapshot of the activity of all the cells. The change in
these profiles in time defines a temporal trajectory in a 3D space.
We observe a divergence between swimming and crawling trajectories
before the behavioral motor pattern begins. This divergence defines
the time at which the choice was made; cells showing activity
differences at this time are good candidates for involvement in
behavioral choice.

Support Contributed By: NSF IGERT Grant (KLB), 
                        La Jolla Interfaces in Science (KLB),  
                        NIH Research Grants MH43396 & NS35336 (WBK)
 
 
Citation:
K.L. Briggman, H. Abarbanel, W.B. Kristan. PRINCIPAL COMPONENT
ANALYSIS OF BEHAVIORAL CHOICE IN THE LEECH CNS. Program No. 605.1.
2003 Abstract Viewer/Itinerary Planner. Washington, DC: Society
for Neuroscience, 2003. Online.