The full understanding of molecular mechanisms of cell differentiation requires a holistic view. Here we combine label-free FTIR and Raman hyperspectral imaging with data mining to detect the molecular cell composition enabling noninvasive monitoring of cell differentiation and identifying biochemical heterogeneity. Mouse adipose-derived mesenchymal stem cells (AD-MSCs) undergoing adipogenesis were followed by Raman and FT-IR imaging, Oil Red, and immunofluorescence. A workflow of the data analysis (IRRSmetrics4stem) was designed to identify spectral predictors of adipogenesis and test machine-learning (ML) methods (hierarchical clustering, PCA, PLSR) for the control of the AD-MSCs differentiation degree. IRRSmetrics4stem provided insights into the chemism of adipogenesis. With single-cell tracking, we established IRRS metrics for lipids, proteins, and DNA variations during AD-MSCs differentiation. The over 90% predictive efficiency of the selected ML methods proved the high sensitivity of the IRRS metrics. Importantly, the IRRS metrics unequivocally recognize a switch from proliferation to differentiation. This study introduced a new bioassay identifying molecular markers indicating molecular transformations and delivering rapid and machine learning-based monitoring of adipogenesis that can be relevant to other differentiation processes. Thus, we introduce a novel, rapid, machine learning-based bioassay to identify molecular markers of adipogenesis. It can be relevant to identification of differentiation-related molecular processes in other cell types, and beyond the cell differentiation including progression of different cellular pathophysiologies reconstituted in vitro.
Keywords: adipogenesis; mesenchymal stem cells; prediction of differentiation stage; regenerative medicine; spectroscopic molecular imaging.
Copyright © 2024 The Authors. Published by Elsevier Inc. All rights reserved.