Circadian Rhythmicity and Timescales in Fly Behavior

Abstract

A quantitative, data-driven assay for classification of behavior in Drosophila melanogaster is applied to day-long imaging datasets to understand the interaction of behavioral timescales over six orders of magnitude, particularly the longest circadian timescale. We show evidence that circadian behavioral modulation exists beyond binary measurements of locomotor activity or inactivity accessible to previous studies. In particular, peaks of activity for different behaviors occur out of phase over the course of the day. Predictability of behavior varies over the course of the day, as do the relative importance of different timescales in explaining behavioral variability. These changes are observed with and without extrinsic circadian cues. We investigate short timescale interactions between behaviors underlying these larger trends. A perspective is developed toward using these analyses to understand circadian regulation of behavioral outputs.