Marmosets are emerging as an important model species for neuroscience research, driven by the development of new technologies such as CRISPR that allow targeted genetic modifications in this species. These developments will allow primate research to take advantage of powerful genetic tools that were previously restricted largely to rodents, including optogenetics, genetic activity reporters, and targeted mutation of endogenous genes implicated in brain function and human disease. Marmosets are well suited to this approach, being small and fast-breeding compared to most primates. They are typically housed in family groups, and exhibit a variety of social behaviors in captivity including complex vocal repertoires. Marmosets thus represent a promising system for studying social behavior and other cognitive functions in a primate model, and they also hold great promise for modeling brain disorders that affect cognitive functions that are difficult to study in other species such as rodents. To take full advantage of these emerging animal models, it is necessary to develop new methods for analyzing their behavior, including naturalistic social interactions that are imperfectly captured by standardized behavioral tasks. We therefore plan to develop a system for automated analysis of marmoset behaviors in the home cage. The system will consist of an integrated array of sensors including video cameras, depth sensors, and collar-mounted wearable microphones. The resulting multimodal data will be synchronized and analyzed using methods from computer vision, speech processing, machine learning, and multimodal data analysis. Specifically we will formulate the tracking analysis as a probabilistic graphical model, which will allow video data to be integrated with audio recordings, and with other modalities that could be explored in future, including inertial motion sensors, physiological recordings and other contextual data. Based on this approach we will develop methods to classify calls, identify individual callers, track the locations and identities of each animal in three dimensions, and classify different actions, including interactions between individuals. We envisage that our system will be useful for a wide range of studies in basic and translational neuroscience, and in particular it will be useful for studying behavioral phenotypes in genetic models of human psychiatric disorders, and for relating behavioral abnormalities to their underlying genetic and neural causes.
We propose to develop an automated system for monitoring the behavior of marmosets in the home cage environment. The proposed system will measure a wide range of naturalistic behaviors including vocal and social interactions, and will be useful for studying animal models of psychiatric disorders such as autism and schizophrenia, in which these behaviors are disrupted. This in turn will allow researchers to relate behavioral abnormalities to their underlying neural and genetic causes, an essential step toward the understanding and eventual treatment of these disorders.
