Machine Learning\nAccelerates Screening of Osteoclast\nDifferentiation Inhibitors from Natural Products

Natural\nproducts that inhibit osteoclast differentiation are promising\ntherapeutic and preventive agents for osteoporosis. Conventionally,\nidentifying osteoclast differentiation involves visual inspection\nof the microscope images of stained osteoclasts. In this study, a\nsupervised machine learning model was developed to classify bright-field\nmicroscope images of osteoclasts without staining. The model was used\nto screen a compound library, and osteoclast differentiation inhibitors\nwere identified, demonstrating the validity of our method. Next, an\nin-house library of fungal extracts was screened, and pinolidoxin\nwas revealed as an inhibitor of osteoclast differentiation. Our machine\nlearning method enabled accurate, objective, and high-throughput evaluation\nof osteoclast differentiation and efficient screening of the inhibitors\nfrom natural product extracts. This study represents the first machine\nlearning classification developed to evaluate the inhibitory activity\nof natural products in osteoclast differentiation.