Answers are grounded in this paper and cite their sources.
Introduction: Label-free profiling of cellular transcriptional and metabolic states during B cell differentiation and malignant transformation remains technically challenging. Raman spectroscopy offers a non-destructive alternative for single-cell biochemical characterization, yet its ability to infer transcriptomic profiles and distinguish leukemic cells has been limited. Methods: We established a Raman spectroscopy-based platform to profile B cell differentiation stages (HSC, pro-B, pre-B, naive B) and leukemic B-ALL cells at the single-cell level. Raman spectra were acquired from fixed cells and analyzed using principal component-linear discriminant analysis (PC-LDA) classification. An adversarial autoencoder (AAE) framework was employed to align Raman measurements with reference single-cell RNA-seq data, generating Raman-inferred transcriptomic profiles in the reference expression space. Cytochrome c expression was validated by flow cytometry, and metabolic pathways were examined via GSEA. Results: PC-LDA resolved four B cell differentiation stages with 96.48% accuracy, identifying Raman features consistent with cytochrome c as potential spectral markers of differentiation status, validated by flow cytometry and mitochondrial membrane potential measurements. The AAE-based model generated Raman-inferred profiles that preserved major cell-type-associated transcriptomic patterns, with a mean classification accuracy of 91.78% ± 8.13% across repeated partitions. Performance varied considerably across repeated splits for pro-B (52.48-100%) and pre-B cells (34.4-100%), indicating that distinguishing closely related stages remains challenging. Applied to B-ALL, the method distinguished cord-blood-derived normal B cells from bone-marrow B-ALL cells (96.62% accuracy) and revealed reprogramming of glucose metabolism, consistent with transcriptomic enrichment analysis. Conclusions: This study presents a proof-of-concept framework demonstrating that Raman spectroscopy, integrated with machine learning and transcriptomic alignment, enables non-destructive, fixed-cell-based omic profiling of B cell development and leukemia. The approach bridges label-free optical readouts with transcriptome-informed profiling, opening new avenues for hematopoietic research, analysis of archived specimens, and future clinical exploration. However, these findings are based on a limited number of donors and unmatched tissue sources, and require validation in larger cohorts before clinical translation.
No takes yet. Share an insight, caveat, or question.
Cheng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: