Participant-level Raman spectroscopy of circulating EV-enriched fractions reveals distributed spectral differences in ischemic stroke patients

Raman spectroscopy provides a label-free vibrational profile of extracellularvesicle (EV)-enriched biofluid fractions. Participant-level Raman profiles were obtained from 97 participants with ischemic stroke and 50 controls. After quality control and region-specific normalization, spectra acquired at individual spatial sampling locations were aggregated into a single median spectrum per participant in the fingerprint (734–1788 cm −1 ) and C–H stretching (2451–3197 cm −1 ) regions, so that inference was carried out at the participant rather than the measurement level. The objectives were to characterize participant-level spectral variation, assess exploratory case-control separation, and examine associations with clinical outcomes within the stroke cohort. Participant-level spectra showed 21 reproducible Raman bands and shoulder features associated with lipids, proteins, nucleic acids, and carotenoids. Lipid-associated regions accounted for most of the normalized spectral variance (fingerprint PC1, 75.8%; C–H PC1, 67.1%), a pattern that persisted when PCA was repeated using only spectra from participants with ischemic stroke. All 21 features differed between stroke and control participants after false-discovery-rate correction, with increases and decreases distributed across multiple correlated Raman domains. The primary whole-spectrum partial least-squares discriminant analysis (PLSDA) model gave modest discrimination (balanced accuracy 0.655, 95% CI 0.538–0.771; ROC–AUC 0.711, 95% CI 0.550–0.823; sex-stratified permutation p = 0.005). Within the stroke cohort, no whole-spectrum association survived correction for multiplicity, and Montreal Cognitive Assessment (MoCA) scores gave negative cross-validated R 2 , so the spectra did not support the prediction of cognitive outcome. Participant-level Raman spectroscopy, therefore, identifies reproducible biochemical differences within compositionally complex EV-enriched fractions and identifies lipid-, protein-, nucleic acid-, and carotenoid-associated regions for molecular follow-up in larger, balanced cohorts.