Ensemble Machine learning model identified citrusinol as functional food candidate for improving myotube differentiation and controlling CT26-Induced myotube atrophy

Skeletal muscle loss leads to decreased quality of life, increased incidence of chronic disease and mortality. To identify functional food materials to alleviate muscle atrophy, we built a multitarget-based machine learning system to identify novel phytochemicals that can inhibit TGF-ss, which induce muscle weakness, and increase PGC-1a, a target of exercise mimetics. The multitarget-based machine learning system is built as an ensemble model of four algorithms with each optimal input representation. Citrusinol was identified by our model, and its anti-atrophy effects were validated using C2C12 cells. Citrusinol enhanced protein synthesis via AKT/mTORC1 pathway, increased myogenic differentiation, and increased PGC-1a and its downstream regulators, MEF2A and TFAM. Citrusinol attenuated CT26-induced myotube atrophy by blocking TGF-ss, p-SMAD3, MAFbx, and TGF ss-induced MuRF1 and p-SMAD3. These results suggest that the proposed model can effectively identify functional foods to manage muscle atrophy; additionally, citrusinol was demonstrated as a promising candidate for future animal experiments.