Machine Learning-Driven Optimization of Therapeutic Substance Composition for High-Hardness, Fast-Dissolving Microneedles for Androgenetic Alopecia Treatment

Abstract

Treating androgenetic alopecia (AGA) with platelet-rich plasma (PRP) holds great promise; however, effective and comfortable delivery remains a challenge. Direct injection causes pain, and PRP-incorporated microneedles (MNs) have low hardness and slow dissolution. To tackle this problem, we propose a machine-learning (ML)-driven strategy, which involves integrating the selection of therapeutic substances, orthogonal experiment designs, ML prediction, and Pareto front identification. Through the implementation of only 18 experiments based on orthogonal experiment designs, this ML-assisted strategy can pinpoint an optimal material composition that concurrently attains high hardness and rapid dissolution. We utilized this optimal material composition to fabricate MNs, and their biological functionality was demonstrated through multiple aspects, including the sustained release of various growth factors over 30 days, more than 90% bacterial inhibition, reactive oxygen species scavenging, and the promotion of the proliferation of dihydrotestosterone-damaged human dermal papilla cells. In vivo studies indicated significant hair regrowth in AGA mice through the activation of the Wnt/β-catenin pathway, outperforming the effects of minoxidil. Significantly, this approach eliminates the biosafety risks associated with the use of synthetic materials. The developed framework is anticipated to serve as a generalizable paradigm for expediting the clinical translation of biomaterials such as MNs.