Pose-derived microstructure resolves how hunger and social isolation shape social–food choice in male mice
Year: 2026
Type: preprint
Abstract: Preference indices reduce choice between competing rewards to a single score, obscuring how internal state and experience shape the behaviour that produces it. We combined the social versus food preference test (SFPT) with markerless pose estimation, cross-validated decoding using markerless pose estimation, and Zif268 immunofluorescence in adult male CD-1 mice that were group-housed (GH) or single-housed (SH) and tested fed ad libitum (AL) or after 16 h of food deprivation (FD). Food deprivation shifted preference toward food in both housing conditions, whereas SH shifted it toward the social stimulus. In GH mice, the food shift reflected longer food-investigation bouts and stronger food-directed orientation, but not more food-chamber entries. SH-FD mice showed no significant preference, yet they investigated both stimuli and approached them at high speed. Pose-derived features decoded feeding state (AUC 0.94) and housing condition (AUC 0.85), whereas manually scored measures decoded feeding state (AUC 0.84) but not housing (AUC 0.56). In exploratory analyses, simultaneous choice between food and social stimuli was associated with higher Zif268 expression in the basolateral and central amygdala and supraoptic oxytocin neurons of GH-FD mice, and in the BNST and hypothalamic oxytocin neurons of SH-AL mice. Pose-derived microstructure thus resolves components of social–food choice that a single preference score merges. Highlights Null preference scores conceal distinct social–food strategies. Food deprivation prolongs food visits, not chamber entries. Single-housed, food-deprived mice show no preference but approach both stimuli fastest Pose features decode housing history; manual scores do not.
Language: English
Subfield: Endocrine and Autonomic Systems
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
Funder: Deutsche Forschungsgemeinschaft
Award: G1162950590
OpenAlex ID: W7234209261
