AUTOGAITA: A VERSATILE QUANTITATIVE FRAMEWORK FOR KINEMATIC ANALYSES ACROSS SPECIES, PERTURBATIONS AND BEHAVIOURS

Distinct behaviours require the nervous system to execute specialised motor programs, i.e., pre-structured patterns of muscle contractions, with characteristic sequence, timing, and force, which lead to specific reflexive or goal-oriented movements, e.g., forward walking, or backward swimming. Importantly, whether the execution and adaptation of these motor programs follow conserved principles across species and perturbations remains unclear. This is partly due to the lack of a standardised tool for such analyses. Although advances in deep-learning methods, such as DeepLabCut and SLEAP, have considerably improved motor assessments by facilitating the tracking of body landmarks across behavioural tasks and species, no universal framework exists to coalesce the generated time-stamped body coordinates into meaningful representations of motor programs.
To provide standardised comparisons of motor programs across species, perturbations and behaviours, we developed the Python toolbox AutoGaitA (Automated Gait Analysis), which generates robust kinematic assessments of any limbed species performing any rhythmic behaviour of interest.
Adopting AutoGaitA for a cross-species walking paradigm, we found that locomotor programs in flies, mice, and humans rely on diverse mechanisms to generate limb propulsive strength, but employ a similar distal-to-proximal gradient of joint movement velocities. Also, we showed that ageing induces a loss of propulsive strength in all species while preserving the velocity gradient. Furthermore, we observed that in mice, locomotor programs adapt as an integrated function of concomitant perturbations, namely ageing and task difficulty. Thus, with AutoGaitA, we began to reveal the conserved and divergent mechanisms underlying the execution of locomotor programs in physiological and perturbed states.