Learning muscle-health trajectories in ageing: towards AI-guided organ chips

Age-related decline in muscle health is commonly defined using cut-offs for strength, muscle quantity or quality, and physical performance. These measures are essential for diagnosis, but prevention and recovery require an understanding of how muscle responds to stress over time. A central question is when recovery remains possible and why older adults with similar clinical measurements follow different trajectories. Clinical cohorts, animal studies and static cell models each address part of this problem. None readily combines human-tissue relevance with controlled perturbation, repeated functional measurement and experimentally testable recovery cues. We propose a clinically anchored workflow that treats sarcopenia as a model problem for studying age-related muscle change as a dynamic process. The sequence begins with a clinical or biological question, applies a standardised perturbation, uses repeated functional readouts, matches analysis to the question and data structure, and validates the resulting interpretation against independent evidence. Human muscle organ chips, particularly when integrated with adipose, immune, vascular or neuromuscular modules, can expose engineered muscle to defined inflammatory, metabolic, unloading or denervation-like stress while tracking functional and molecular responses. Current evidence remains largely proof of concept and does not support prediction of patient-specific recovery windows, responder status or long-term sarcopenia progression. AI should support this experimental sequence by organising time-course data, quantifying donor-, platform- and measurement-related uncertainty, prioritising informative readouts, and suggesting follow-up experiments. Within these limits, AI-assisted organ-chip studies could help investigate reversible decline, tissue crosstalk and divergent recovery in ageing.