High-throughput, label-free monitoring of cellular differentiation remains a major challenge in stem cell biology and regenerative medicine. Raman spectroscopy offers rich molecular specificity without perturbing the cell state, but the analytical complexity of large, unlabeled spectral data sets has limited its adoption. Here, we introduce a scalable computational framework that adapts algorithms from single-cell genomics for the analysis of line-illumination Raman spectroscopy data. Applying this approach, we track the stepwise differentiation of human induced pluripotent stem cells into hepatocyte-like cells at single-cell resolution across more than 1.8 million spectra. By integration of unsupervised clustering with supervised learning, our pipeline enables rapid analysis (<2 min per imaging field), monitoring key biochemical markers, such as cytochromes, glycogen, and lipids, and real-time discrimination of successful and aberrant differentiation without labeling. This work establishes a generalizable strategy for Raman-based cell state profiling and supports non-invasive, in-line monitoring in stem cell manufacturing pipelines.
Raj et al. (Thu,) studied this question.