Evaluation of Hepatic Progenitor and Hepatocyte-Like Cell Differentiation Using Machine Learning Analysis-Assisted Surface-Enhanced Raman Spectroscopy.

Abstract

Technology has been developed to monitor the differentiation process of human mesenchymal stem cells (hMSCs) into hepatocyte-like cells (HLCs) and hepatic progenitor cells (HPCs). These cell lineages, differentiated from MSCs, are ethically unproblematic and are gaining attention as promising cell-based therapies for treating various liver injuries. High-sensitivity, label-free, real-time monitoring technologies integrated with artificial intelligence have been used to evaluate and optimize cell differentiation for enhancing the efficiency of cell therapy delivery. Using an Au-ZnO nanorod array-based surface-enhanced Raman scattering (SERS) sensing chip, cell differentiation from hMSCs to HPCs and HLCs was nondestructively monitored through spectral analysis of cell secretions. Principal component extraction was employed to reduce variables, followed by discriminant analysis (DA). The application of principal component-linear discriminant analysis (PC-LDA), an artificial intelligence algorithm, to spectral data enabled clear grouping of hMSCs, HPCs, and HLCs, with monitoring accuracies of 96.3%, 98.8%, and 98.8%, respectively. Spectral changes observed during the differentiation from hMSCs to HPCs and from HPCs to HLCs over several days demonstrated the effectiveness of SERS combined with machine learning algorithm analysis for differentiation monitoring. This approach enabled real-time, nondestructive observation of cell differentiation with minimal sample labeling and preprocessing, making it useful for sensing differentiation validation and stability. The machine learning- and nanostructure-based SERS evaluation system was applied to the differentiation of ethically sourced MSCs and demonstrated substantial potential for clinical applicability through the use of patient-derived samples.