Computer vision-aided locomotor behavioral analysis identifies therapeutic motor signatures in a mouse model of Huntington disease

Abstract

Huntington disease (HD) is a neurodegenerative disorder characterized by progressive motor dysfunction. Traditional open-field tests quantify spontaneous locomotor parameters; however, fine mouse motor signatures, particularly disease stage-specific changes in HD motor symptoms and pharmacodynamic responses to therapeutic treatments. Here, we employed a computer vision-aided behavioral flow analysis designed to quantify fine, HD-relevant motor dysfunction in the zQ175DN HD mouse model, ranging from early HD-like motor signatures to well-defined motor deficits. Markerless pose estimation and Keypoint-MoSeq segmented standard top-view open-field recordings into recurrent behavioral syllables, which were then organized into higher-order clusters and transition networks. Disease stage-dependent changes in syllable occurrence, syllable duration, behavioral-state composition, and transition structure were identified. These analyses are not possible with traditional open-field assays. Syllable-duration features provided the strongest genotype discrimination, and HD-like motor features were also characterized by hub remodeling and transition-network disorganization. These features were integrated into an HD motor dysfunction (HDMD) score based on age- or HD progress-matched wild-type (WT) -standardized absolute deviations. The HDMD score distinguished HD mice from WT across multiple symptomatic stages and correlated with HD pathology and disease severity. Effect-size and power analyses suggested improved efficiency for detecting potential therapeutic effects. This framework requires only standard top-view recordings and may also support retrospective analysis of existing open-field video datasets. Overall, the HDMD framework provides a practical strategy for identifying fine motor changes in HD mice, aiding study design and preclinical efficacy assessment in HD drug development.