bioRxiv Science⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.03.23.713050

Integrating supervised and unsupervised machine learning for behavior segmentation reveals latent frailty signatures and improves aging clocks in isogenic and outbred mice

Sabnis, G.·Miao, D. M.·Kumar, V.
View original source ↗Export metadata ↓

Abstract

1Manual frailty index (FI) assessment in mice is a strong predictor of morbidity and mortality, and is frequently used in mechanistic and translational geroscience. However, it is labor-intensive, requires expert training, and is vulnerable to scorer variability. We previously developed a visual frailty index (vFI) that objectively predicts age and frailty using expert-defined, supervised behavioral features extracted from open-field videos. However, relying solely on human-defined features may miss subtle, latent behavioral signatures of aging. Here, we test whether unsupervised behavioral discovery using Keypoint-MoSeq (KPMS) could uncover these hidden signatures and improve the prediction of aging-related outcomes. Using a large dataset of isogenic C57BL/6J (B6J) and genetically diverse Diversity Outbred (DO) mice, we find that unsupervised features are highly predictive of chronological age, biological frailty, and the proportion of life lived. Notably, while supervised features overall outperformed unsupervised features in these tasks, combining both feature sets yielded the highest predictive accuracy across all outcomes. Despite these improvements, models trained on either feature set failed to generalize across strains, confirming that behavioral manifestations of aging are strongly population-specific. These findings demonstrate that supervised and unsupervised machine vision provide complementary information, establishing a highly sensitive, scalable, and non-invasive framework for objective and scalable geroscience in rodents.

Explore related subjects

animal behavior and cognitionPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Sabnis, G., Miao, D. M., Kumar, V.. 2026-03-25. Integrating supervised and unsupervised machine learning for behavior segmentation reveals latent frailty signatures and improves aging clocks in isogenic and outbred mice. https://doi.org/10.64898/2026.03.23.713050

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

A comparison of female competitive traits: Female aggression peaks at nest building but female song spans multiple contexts in a temperate songbird

Female-female competition is increasingly recognized as a key driver of female ornamentation, including birdsong, which often functions in intrasexual competition. However, the specific resources females use elaborate traits to compete for remain unclear. In addition, few studies have simultaneously investigated the use of multiple competitive traits in females, despite growing independent interest in these traits (e.g., female song and aggression). We investigated the competitive contexts of female song, aggression and calling behavior in northern house wrens (Troglodytes aedon) to determine which resources females compete for across the breeding season. We simulated conspecific territorial intrusions using female song at three breeding stages representing different contexts: arrival (mate and territory acquisition), nest building (nest site and breeding status defense), and egg laying (brood defense). We tested whether female song and physical aggression varied as reproductive resources shifted across the breeding cycle. Females were significantly more aggressive during nest building, showing 5.8 times greater odds of a higher-intensity aggressive response during nest building compared to arrival. Female song output was similar across early stages but declined during egg laying, though this was not statistically significant after correction for multiple comparisons and individuals varied substantially in overall singing propensity. Non-song vocalizations varied by call type and breeding stage. Calls associated with aggression occurred most frequently during nest building, consistent with peak physical aggression responses. Together, these results identify nest building as the stage of highest female aggression, consistent with heightened competition over nest cavities and associated breeding status in this cavity-nesting species. In contrast, female song occurred across all stages and appears to function in multiple competitive contexts. This study provides evidence for context and mode-specific female signaling in a temperate songbird and highlights that females strategically use aggression, calls, and song to mediate social conflict across breeding contexts.

animal behavior and cognition↗
bioRxiv2026

Tracking human foragers and their prey reveals adaptive predator-prey dynamics

Hunting for mobile prey is thought to have played a key role in hominin evolution, by providing high-quality nutrition that supported the development of the exceptionally large human brain. However, human-prey dynamics remain poorly understood because studies have not yet tracked human foragers and their prey simultaneously. Here, we employ high resolution tracking of groups of human foragers (ice-fishers) and their prey (fish shoals) to study human-prey dynamics. Our results show that foragers adaptively combined personal and social information in deciding where to forage and for how long, closely matching the prey distribution. Prey responded dynamically to human exploitation, showing increased attraction to fishing activity, alongside decreased biting probability. Furthermore, we found that foragers adaptively relied on memory, preferentially returning to areas with high prey presence, particularly when their current return rate was low. Our results show how human foragers overcome the challenges of extracting invisible, mobile and reactive prey by tightly tuning patch-selection, patch-leaving and patch-return decisions to the distribution and behaviour of their prey.

animal behavior and cognition↗
bioRxiv2026

4-Dimensional Chess: Acoustic Localisation Reveals Nested Spatio-temporal Strategies in an Arboreal Communication Network

1. Adaptive behavioural strategies require animals to simultaneously navigate social and ecological domains across multiple spatial and temporal scales. Although drones and computer vision have recently transformed the study of wild animal societies, many nocturnal species and those occupying structurally complex habitats remain inaccessible to these approaches, limiting our understanding of behaviour in natural settings. 2. We aimed to determine how behavioural strategies are organised across nested spatial and temporal scales within a wild communication network. 3. We used three-dimensional acoustic localisation and source separation to track individual male Hyperolius sp. A, a nocturnal African reed frog, within a natural rainforest chorus and quantify patterns of site fidelity, movement, spatial organisation, and call-timing interactions. 4. Males exhibited significant site fidelity across nights, while chorus spatial structure varied with local caller density. Within nights, individuals followed a stereotyped behavioural sequence, descending from elevated arboreal refugia before settling into lower calling positions near breeding sites. At finer temporal scales, call-timing interactions varied according to both local competitor density and the proximity of neighbouring rivals. 5. These findings demonstrate that behavioural strategies emerge across nested spatial and temporal scales and that long-term spatial positioning, short-term movement decisions, and moment-to-moment signalling interactions are tightly linked within natural communication networks. More broadly, acoustic localisation provides a powerful framework for studying behaviour in species and habitats that remain difficult to observe using conventional approaches.

animal behavior and cognition↗