Establishing the unified, automatic, unbiased platform for characterization of social behavior in mice

Validation of any biological intervention developed for tackling the root or symptoms of neuropsychiatric disorders requires proper tools. Disturbances of social behavior is a hallmark of many psychiatric conditions, such as depression, anxiety and schizophrenia. Capturing and quantifying broad range of social behavior simultaneously within a single tool is therefore essential, as it would allow for registration and clustering of groups of behavior specific for particular disorders. Such method would be of great advantage for better translational studies.
Here, within Same-NeuroID project, we implemented a pipeline for efficient characterization of complex social behavior in mice. We call it Same-SocialBox. In the setup animals are housed together in a semi-naturalistic environment with proper bedding, food and water access, and with night-day cycle. Animals are recorded for long hours to ensure a capture of diverse behavior. SLEAP is used for pose estimation of recorded animals and extracted coordinates are processed by deepOF software to create set of features. These features are then used to “feed” models of both supervised and unsupervised learning for complex behavior classification.