Molecular Profiling of Inflammatory and Myofibroblast Cancer-Associated Fibroblast Subtypes Derived from Human Pancreatic Stellate Cells Using Machine Learning-Based Label-Free Raman Spectroscopy.

Cancer-associated fibroblasts (CAFs), one of the most substantial constituents of the pancreatic tumor microenvironment, exhibit far greater heterogeneity and phenotypic plasticity than it was previously recognized. Accordingly, distinguishing between CAF subpopulations and their functional roles in pancreatic tumorigenesis has become increasingly important. Additionally, as the importance of the therapeutic approach increases, interests in technologies capable of efficiently differentiating between normal fibroblast subpopulations and pathologic CAFs also grow. Label-free imaging and analytical technologies that do not require fluorescent labeling or other preprocessing steps offer a promising alternative to conventional invasive cell analysis. Here, we employed Raman spectroscopy to chemically characterize human primary pancreas stellate cell (HPaSC), inflammatory CAF (iCAF), and myofibroblastic CAF (myCAF) derived from HPaSC at the cellular level for molecular profiling. As a result, we successfully compared the distinctive biological and chemical properties of each fibroblastic subtype. These Raman spectrum findings were validated by transcriptomic and lipidomic analysis. Our molecular profiling demonstrates that CAF subpopulations can be quantitatively distinguished based on their intrinsic chemical signatures, offering valuable insights into identifying and characterizing CAFs without relying on fluorescence or specific biomarkers. These multivariate spectral analyses enable subtype classification in 95% accuracy combined with partial least squares discriminant analysis (PLS-DA). This result demonstrates that CAF subtypes can be quantitatively distinguished using their intrinsic molecular signature, which support potential in pancreatic cancer research and therapeutic development.