Behavioral phenotypes in aging: structured exploratory computational analysis of multi-assay behavioral data

Aging is associated with progressive alterations in behavioral function, yet majority of the studies interpret behavioral outcomes on an assay-by-assay basis, limiting understanding of how behavioral domains are organized. Our previously published study used a mouse cohort to report assay-specific behavioral effects. In this study, we examined whether aging-related behavioral signals remain localized to individual assays or can be summarized as coordinated domain-level patterns.

A structured rule-based workflow was implemented for feature engineering, direction harmonization, z-score standardization, and domain-level composite construction across locomotion/exploration, anxiety/avoidance, depression/passive coping, cognition/learning, memory, and sociability domains. Outcomes were analyzed using parametric or non-parametric models following assumption screening, with effect sizes reported.

Domain-level composite scores did not show significant age- or sex-related effects, indicating limited broad behavioral separation. In contrast, refined feature-level analysis identified modest locomotor differences (core locomotion distance; Kruskal-Wallis p = 0.048) and the clearest age-related signal in Barnes Maze performance, particularly the Barnes Maze efficiency index (F = 10.815, p < 0.001), with reduced performance in older animals. Repeated-measures analyses further confirmed training-related improvements in latency across days. Several additional measures showed a trend but did not reach statistical significance and hence are reported descriptively only.

Aging-related behavioral changes in this dataset were concentrated in specific assay-level measures rather than broadly distributed across domains. Domain-level aggregation reduced separation of effects, indicating that the present composites should be interpreted as heuristic summaries rather than validated behavioral dimensions. The main added value of this reanalysis is therefore interpretive, showing that the strongest signals remain most evident in selected measures, particularly Barnes Maze outcomes.