Machine learning-informed digital measures improve the sensitivity and translational relevance of in vivo neurodegenerative disease models

Neurodegenerative diseases present significant global health challenges, with Amyotrophic Lateral Sclerosis (ALS) being a particularly devastating disorder. ALS leads to motor neuron degeneration, muscle weakness, and loss of motor control, with an average life expectancy of just 2-5 years post-diagnosis. Developing sensitive, translationally relevant preclinical measures is critical for understanding disease mechanism and advancing therapeutic strategies. Conventional in vivo assessments of neurodegeneration are often subjective, labor-intensive, and lack temporal resolution. Advances in sensor technology and machine learning offer new opportunities to enhance preclinical research. Digital measures of behavior and physiology provide continuous, objective, and biologically relevant insights into the disease progression and treatment effects. Through the Digital In Vivo Alliance (DIVA), we evaluated machine-learning-informed digital measures in the B6.Cg SOD1-G93A ALS model using home-cage computer vision cameras with infrared detection. We hypothesized that digital measures would provide greater sensitivity than standard measures. Male SOD1 and non-carrier (NCAR) mice were continuously monitored from 4 through 26 weeks of age via the JAX Envision TM platform alongside standard measures. Progressive neuromuscular deficits were observed, including increased neuroscores (>1 at 19 weeks), as well as declining compound muscle action potential (CMAP) and greater signal reduction in repetitive nerve stimulation (RNS) at 12 and 16 weeks. Serum neurofilament light chain (NFL) was elevated at these time points. Digital measures revealed declining movement near the study endpoint, with a breakpoint detector generated through segmented line regression (median breakpoint: 19.1 weeks) outperforming linear models (median R 2 improvement: 0.0995). Sleep disturbances emerged by 6 weeks, progressing to fragmented sleep after 14 weeks. Digital measures provided greater sensitivity in detection of disease onset earlier than standard measures. This innovative approach demonstrates the value of objective, longitudinal digital measures in preclinical ALS research. By improving sensitivity and translational relevance, these approaches could accelerate the development of effective interventions and address a critical unmet need in neurodegenerative disease research. This work was funded by the Digital In Vivo Alliance. This abstract was presented at the American Physiology Summit 2025 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.