If we could take a picture of a memory, what would it look like? How do you store the memory of a fragrance, or of how to ride a bicycle?
We are working on the idea that memories are stored in the pattern of connections between brain cells. To store memories in this manner, the brain must convert information into activity patterns, modify connections,
and stably retain these modifications. Our research covers each of these topics.
First, the data must be converted from sensory
input (like odor molecules) into patterns of activity in the brain. This
'representation' of information sets the stage for subsequent computations
and memory storage. To use an analogy, a computer
represents sound as a sequence of ones and zeroes, and once it is in
that form it can be played back as music, or stored as an mp3 file. To tap
into representation of odors in the brain, we record the timing and
location of activity across brain cells when animals are exposed to odors
in different contexts. We train rats to track odors to their source, or to
push paddles depending on odor input, and see how brain activity relates to
these tasks. Details
Second, the brain sets up connections between cells depending on input. To
work out if two cells are connected, we test if signals from the first cell
can affect activity in the second. Building on advances in microscopy, we
can monitor single-cell activity using chemical sensors whose light emission
changes when a cell is active. Thus, in a section of brain, the brightly
lit-up cells represent the ones which have received input. By systematically
stimulating different inputs, we can build up small but precise connection
diagrams. We do these recordings in the rat hippocampus, a region of the brain
involved in memory. We anticipate that as these connection diagrams scale up,
we will begin to see traces of memory storage in the connection patterns. Details
Third, the connections need to be stable to store information for a long time.
This is very hard to do, because each connection, or synapse, is so small that
a relatively small number of individual molecules must do all the work and
withstand thermal noise, turnover, and chemical insults. We
probe events at these tiny scales using both experiments and computer models.
In the computer models we analyze how tiny molecular circuits in each
synapse can do computations and can store information reliably. This turns
out to be closely coupled to many other cellular processes: the electrical
signals in brain cells, physical reshaping of synapses, and synthesis of
new proteins. We are developing powerful software tools to model how
these events are orchestrated. This software, MOOSE, runs from windows laptops
to giant Unix-based supercomputers. Details
Overall, our work falls into the domains of systems biology and computational
neuroscience, with a lively mix of experiments and computer modeling. Our lab
includes people from physics, chemistry, mathematics, biology,
computer science and other branches of engineering.
Dr. Tara Thiagarajan is a Visiting Scientist at NCBS, working on coherence potentials in the cortex. She is hosted by the Bhalla lab.
