High-throughput unsupervised quantification of patterns in the natural behavior of marmosets

Abstract

Recent advances in genetic engineering have accelerated the production of nonhuman primate models for neuropsychiatric disorders. To use these models for preclinical drug testing, behavioral screening methods will be necessary to determine how the model animals deviate from controls, and whether treatments can restore typical patterns of behavior. In this study, we collected a multimodal dataset from a large cohort of marmoset monkeys and described typical patterns in their natural behavior. We found that these behavioral measurements varied substantially across days, and that behavioral state usage was highly correlated to the behavior of cagemates and to the vocalization rate of other animals in the colony. To elicit acute behavioral responses, we presented animals with a panel of stimuli including novel, appetitive, neutral, aversive, and social stimuli. By comparing these behavioral conditions, we demonstrate that outlier detection can be used to identify atypical responses to a range of stimuli. This data will help guide the study of marmosets as models for neuropsychiatric disorders.