RamanOmics Maps the Biochemical Fingerprint of Aging Cells in Living Tissue

Understanding aging requires more than gene expression data. Lipid composition, protein conformation, extracellular matrix chemistry, and metabolite states all govern cell function—yet these biochemical layers are invisible to standard genomic tools. RamanOmics addresses this gap by combining single-nucleus RNA sequencing (snRNA-seq), STARmap in situ sequencing and hybridization (STARmap-ISS/ISH), and label-free hyperspectral Raman imaging into a single multimodal platform that maps vibrational-biochemical and molecular states at single-cell resolution within intact tissues.

The team collected lung and skin tissue from young (2-month) and old (26-month) mice, generating over 47,000 high-quality single-nucleus transcriptomes alongside spatially registered Raman images and spatial transcriptomic maps covering 890 genes including 190 senescence-specific markers. Tissue-level aging signatures were strikingly tissue-specific: old lungs showed chronic immune activation, endothelial and vascular remodeling, and ECM dysregulation, while old skin exhibited metabolic decline, impaired ion homeostasis, and altered RNA-processing programs. Earth Mover's Distance analysis quantified transcriptional divergence at the cell-type level, pinpointing endothelial cells and T cells in lung and fibroblasts and interfollicular epidermal cells in skin as the most age-altered populations.

Using p21 as a canonical senescence marker, the authors identified senescent cell populations and their organ-specific transcriptional programs. Lung senescent cells were dominated by ECM remodeling and TGF-β signaling genes (Serpine1, Dab2, Igfbp7), while skin senescent cells featured keratinization and barrier homeostasis modules (Krt10, Lor, Sbsn). Crucially, Raman imaging revealed a conserved biochemical signature across both tissues: a spectral peak at 1131–1135 cm⁻¹ linked to branched-chain fatty acids that robustly distinguishes p21+ senescent cells from non-senescent neighbors without any labeling or tissue destruction.

To unify these modalities, the researchers developed a machine learning-derived 'multimodal barcode' that quantitatively integrates Raman spectral features with transcriptional data. This barcode enables non-destructive in situ identification of senescent cells and provides a richer, more comprehensive readout of cellular state than either modality alone. In a wound-healing experiment, RamanOmics captured dynamic rewiring of senescent cells, documenting coordinated upregulation of barrier-repair genes (Krt10, Lor, Sbsn, Sfn, Dmkn) alongside corresponding increases in lipid-associated Raman signals—confirming that the platform can track biological state transitions in response to physiological stimuli.

As a preprint, these findings await peer review. The study is currently limited to mouse models, and translation to human tissues will require validation. Raman imaging throughput and integration with existing clinical pathology workflows remain practical challenges. Nevertheless, RamanOmics represents a conceptual and technical advance toward non-destructive, label-free profiling of cellular aging states in living or minimally processed tissues.