Meeting Abstract

S1-1.4  Wednesday, Jan. 4  A data-driven methodology for analyzing the behavior of terrestrial fruit flies BERMAN, Gordon J*; BIALEK, William; SHAEVITZ, Joshua W; Princeton University; Princeton University; Princeton University

The last decades have seen an explosion in our ability to characterize the molecular, cellular and genetic building blocks of life; the ingredients out of which we try to explain the rich and compelling behavior of living organisms. Our characterization of behavior itself, however, has advanced more slowly. Since modern ethology was founded over a century ago, behavioral experiments have focused largely on a restricted set of behaviors within the scope of a limited environment. Moreover, the set of behaviors to be examined is often user-defined, creating the potential for human bias and anthropomorphism. The research presented here describes a data-driven methodology for analyzing animal behavior, focusing on the fruit fly, Drosophila melanogaster, as a model system. Towards this end, we have built an imaging system that can track single flies as they move about a relatively unencumbered environment. Utilizing this capacity to generate large data sets of animal behavior, we have developed a method for automatically identifying behavioral states using techniques from image analysis, machine learning, and nonlinear dynamics. Identifying these states provides the starting point for many analyses and creates the possibility for automatic phenotyping of subtle behavioral traits.