|
Résumé |
Behavior and decision-making are determined by physical processes taking place in
the complex environment of the brain. Experimental techniques have reached the
point where it is now possible to map the complete wiring diagram (the physical
connectome of synaptic connections between neurons) of the brain of simple model
organisms at the level of single synapses, and to manipulate individual neuron
activity in freely moving animals and observe the resulting behavior. Together
this lets us investigate how the structure of the connectome constrains an
organism’s capability to process information and generate behavior. I will discuss
how we can combine knowledge of an animal’s connectome with large-scale behavioral
experiments to link neural circuits to decision making and specific behavioral
sequences. I will mainly focus on how to extract the statistical regularities
(i.e., “motifs”) of a connectome of an animal. Relying on a restricted set of
statistically regular circuit motifs, optimized for specific functions, may
provide an animal with biologically advantageous inductive biases for efficient
learning and help encode innate behaviors. Information theory furthermore tells us
that the presence of statistically regularities would make the connectome
compressible, and circuit motifs would thus provide a means to encode the neural
wiring information in the limited storage space of the genome. Identifying motifs
in a connectome is a challenging inverse problem since we have access to only a
single experimental realization (i.e., a single graph). To circumvent problems
with classic null-model-based analysis linked to multiple testing and the ill-
posed problem of defining a proper null model against which statistical
significance is defined, we have developed methods combining hierarchies of
microcanonical random graph null models and graph compression techniques. We
applied our methods to uncover circuit motifs in different brain regions of adult
and larval Drosophila as well as C. elegans in different developmental stages. Our
preliminary results show more compressible yet more complex brain structure in
larger brains. By comparing typical circuit structures in different brains regions
and animals, we may furthermore formulate hypotheses linking circuit structure to
function that can be tested in behavioral experiments. I will finally discuss how
we can build generative models of neural connectomes. |
