Classification of social behavioral responses in stress and non-stress adult male mice with high precision

Abstract To better characterize social behavior following stress exposure in mice, this study introduces a bidimensional analytical framework that extends beyond the limits of conventional analysis. By integrating a time-based social interaction ratio with the average distance to the CD1 aggressor, we propose a composite index that offers a more comprehensive assessment of social engagement during the social interaction test. This metric distinguishes between socially hesitant male and female mice— those entering the interaction zone while maintaining a relative distance from the CD1 aggressor—and mice that display robust sociability by both entering the zone and closely approaching the aggressor. By treating distance as a continuous variable, this approach moves beyond binary zone-based measures and enables a more refined phenotyping of individual differences along the resilience–susceptibility spectrum. This advancement is made possible using open-source, multipose-estimation tools such as DeepLabCut and DeepOF, which allow for high-resolution tracking and behavioral quantification. Our framework refines current preclinical models by capturing subtle behavioral adaptations to social stress. This ultimately improves their translational value for studying the neural and behavioral correlates of stress-related psychiatric disorders.