Search

To understand the relationship between neural anatomy and behavior, the ultimate output of the nervous system, we performed a high-throughput, thermogenetic screen of 2,200 transgenic Drosophila lines from the Janelia GAL4 collection. Each GAL4 line drives expression in a different, sparse subset of neurons in the fly nervous system. Using genetic techniques, we selectively activated these sparse subsets of neurons, and measured the behavioral effects. We developed a high-throughput, automated system for measuring the flies’ locomotion and social behavior with breadth and depth. We recorded 20,000 videos of groups of flies freely behaving in an open-field walking arena, totaling ∼400 TB of raw data. From the video, we tracked the flies’ body and wing positions using our tracking software, Ctrax. We used our machine learning-based behavior classification system, JAABA, to create 15 behavior classifiers (e.g. walking, chasing) that input trajectories from Ctrax and output predictions for each frame of each fly's behaviors (totaling ∼175 billion annotations of behavior). For each line of flies, we computed ∼200 behavior statistics, such as the fraction of time spent chasing, or average speed while walking, summarizing the behavioral effects of activating the targeted neurons in a concise, interpretable manner. Concurrent with our screen, the Janelia Fly Light project imaged the expression pattern of each GAL4 line, producing image stacks indicating which neurons are likely being activated in each line. By jointly analyzing these behavior and anatomy measurements, we created brain-behavior maps suggesting neural substrates for the generation and/or regulation of each behavior. This study is groundbreaking on two technological fronts. It represents the application of video-based behavior recognition to a highly phenotypically diverse data set of thousands of Drosophila genotypes, and the largest unbiased screen to assign function to neurons throughout the nervous system.