Machine learning framework for passage-based quality control of Hanwoo muscle satellite cells for cultured meat

Abstract

Myogenic satellite cells from Hanwoo (Korean native cattle) are valuable for cultured meat and livestock biotechnology, serving as a primary cellular substrate for the biomanufacturing of alternative protein, particularly cultured beef, and enabling the development of standardized, high-value cultured beef products. We established a multivariate grading framework that integrates indicators of proliferation and senescence to objectively assess passage-dependent cellular quality. Missing biomarker values (G0/G1 phase, early apoptosis, γH2AX, SA-β-galactosidase, telomerase) were imputed using live cell number, with random forest regression showing superior performance (R 2 > 0.93). Principal component and K-means analyses (k = 3) identified three grades, early (P2-P3), intermediate (P4-P7), and late (P8-P12), which were further confirmed by partial least squares discriminant analysis. SA-β-galactosidase, live cell number, and telomerase were the most influential discriminators. RNA sequencing validated progressive transcriptional shifts from proliferation to senescence across the passage. This framework supports the establishment of standardized lot-release criteria and scientifically justified decisions regarding passage limits in cultured beef production, thereby strengthening technological readiness and enabling regulatory-aligned quality-control strategies essential for scalable alternative-protein manufacturing. Although this study was conducted using cells derived from a single donor, future validation across multiple donors will be required to assess the robustness and generalizability of this framework.