RamanOmics Decodes Spatial Vibrational-Molecular Architecture and Rewiring in Aging and Repair
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (Arcadia Science)
Abstract
Aging and tissue repair involve multilayered and spatially heterogeneous remodeling across transcriptional, biochemical, and cellular dimensions, yet prevailing definitions rely on isolated molecular markers that obscure how biochemical and transcriptional states co-evolve in tissues. Here we present RamanOmics, a multimodal framework that integrates single-nucleus RNA sequencing (snRNA-seq), spatial transcriptomics, and label-free Raman imaging to map the spatial vibrational–biochemical and molecular architecture of aging and senescence directly in intact tissues. Applied to mouse lung and skin, RamanOmics generates spatially resolved biochemical–molecular maps revealing tissue-specific programs: lung senescent cells are enriched for extracellular matrix (ECM) remodeling and TGF-β signaling ( Serpine1, Dab2, Igfbp7 ), whereas skin senescence is dominated by keratinization and barrier homeostasis modules ( Krt10, Lor, Sbsn ). Across tissues, we identify a conserved branched-chain fatty-acid-linked biochemical profile and Raman signature (1131-1135 cm⁻¹) that robustly marks p21 ⁺ senescent cells. To unify these layers, we develop a machine learning derived “multimodal barcode” that quantitatively integrates biochemical and transcriptional features, enabling non-destructive identification of senescence in situ . In a wound-healing model, RamanOmics further reveals coordinated reactivation of barrier-repair programs in senescent cells, marked by upregulation of Krt10 , Lor , Sbsn , Sfn , and Dmkn together with matching increases in lipid-associated Raman signatures, confirming biological generalizability beyond steady-state aging. By directly integrating gene programs to spatial vibrational–biochemical states, RamanOmics provides a general framework and resource for scalable, multimodal profiling of cellular states.
Article activity feed
-
Second, p21+ senescent cells remain a rare population even in old tissues, limiting the number of cells available for training and potentially reducing classifier power. This highlights the need for high-speed Raman microscopy, larger multi-tissue datasets, and integration of consensus senescence signatures.
Good call out!
-
Extended Data Fig.9
Minor point- I'm finding this difficult to read, especially the tables.
-
hyperspectral Raman imaging (600-1800 cm⁻¹, 873 dimensions with a pixel size of 3 µm)
How did you account for spatial mixing that may occur with the given analyzed spot size? It's possible that a neighboring cell signal could be contributing to the target cell.
-
we manually selected corresponding cellular keypoints across both imaging modalities. This selection tool then generated a 3×3 transformation matrix to adjust the STARmap images to align with the Raman regions. The manual alignment process utilized a least-squares method, employing a modified two-dimensional version of Horn’s (1987) algorithm to account for differences in translation, scale, rotation, and reflection. For each Raman-STARmap paired sample, hundreds of keypoints were manually selected, and the fitgeotrans function in MATLAB was used to transform the STARmap image to match the Raman region. The imshowpair function was employed iteratively after every 20 keypoints to ensure satisfactory alignment.
I appreciate that this is an important and sometimes tricky problem, but well worth doing! Did you consider using an accuracy …
we manually selected corresponding cellular keypoints across both imaging modalities. This selection tool then generated a 3×3 transformation matrix to adjust the STARmap images to align with the Raman regions. The manual alignment process utilized a least-squares method, employing a modified two-dimensional version of Horn’s (1987) algorithm to account for differences in translation, scale, rotation, and reflection. For each Raman-STARmap paired sample, hundreds of keypoints were manually selected, and the fitgeotrans function in MATLAB was used to transform the STARmap image to match the Raman region. The imshowpair function was employed iteratively after every 20 keypoints to ensure satisfactory alignment.
I appreciate that this is an important and sometimes tricky problem, but well worth doing! Did you consider using an accuracy metric for the registration?
-
In mouse skin, elevated Raman peak clusters at 1112-1141 cm⁻¹ (e.g., 1128 cm⁻¹, myristic acid; 1130 cm⁻¹, 12-methyltetradecanoic acid (15Aiso); 1131 cm⁻¹, palmitic acid and other fatty acids; 1134 cm⁻¹, 13-methylmyristic acid (15iso); 1135 cm⁻¹, 15-methylpalmitic acid (17iso)) and at 1434-1444 cm⁻¹ (e.g., 1438 cm⁻¹, palmitic acid; 1439 cm⁻¹, vaccenic acid; 1440 cm⁻¹, oleic acid), coinciding with transcriptomic downregulation of genes involved in ECM (Postn) and muscle contraction (Tnnt3, Ttn)86,87 and upregulation of keratinization genes (Sbsn, Lor)88 (Fig. 5d). In contrast, decreased peaks at 933-948 cm⁻¹ (e.g., 934 cm⁻¹, D-(+)-mannose; 940 cm⁻¹, amylopectin) and 1161-1165 cm⁻¹ (e.g., 1161 cm⁻¹, quinoid ring deformation; 1162 cm⁻¹, adenine) were associated with downregulation of DNA damage response gene Eepd1, and upregulation of skin …
In mouse skin, elevated Raman peak clusters at 1112-1141 cm⁻¹ (e.g., 1128 cm⁻¹, myristic acid; 1130 cm⁻¹, 12-methyltetradecanoic acid (15Aiso); 1131 cm⁻¹, palmitic acid and other fatty acids; 1134 cm⁻¹, 13-methylmyristic acid (15iso); 1135 cm⁻¹, 15-methylpalmitic acid (17iso)) and at 1434-1444 cm⁻¹ (e.g., 1438 cm⁻¹, palmitic acid; 1439 cm⁻¹, vaccenic acid; 1440 cm⁻¹, oleic acid), coinciding with transcriptomic downregulation of genes involved in ECM (Postn) and muscle contraction (Tnnt3, Ttn)86,87 and upregulation of keratinization genes (Sbsn, Lor)88 (Fig. 5d). In contrast, decreased peaks at 933-948 cm⁻¹ (e.g., 934 cm⁻¹, D-(+)-mannose; 940 cm⁻¹, amylopectin) and 1161-1165 cm⁻¹ (e.g., 1161 cm⁻¹, quinoid ring deformation; 1162 cm⁻¹, adenine) were associated with downregulation of DNA damage response gene Eepd1, and upregulation of skin barrier-maintenance genes (Sfn, Krt10).
Similar comment here regarding overinterpretation of the Raman peaks. Unless you use another complementary technique that is truly capable of molecular structure/identity characterization, this is an overstatement. However, I think your overall point still stands regarding lipid-associated C-C modes as relevant signatures and markers.
-
In mouse lung, we identified distinct Raman peak clusters, including 602-630 cm⁻¹ (607 cm⁻¹, glycerol; 614 cm⁻¹, cholesterol ester), 1106-1111 cm⁻¹ (1108 cm⁻¹, α-D-glucose; 1109 cm⁻¹, amylopectin), and 1128-1135 cm⁻¹ (1130 cm⁻¹, 12-methyl-tetradecanoic acid (15Aiso); 1131 cm⁻¹, palmitic acid; 1134 cm⁻¹, 13-methylmyristic acid (15iso); 1135 cm⁻¹, and 15-methylpalmitic acid (17iso)).
Could you explain how you are justifying the molecular assignments here? I understand these peaks can be tied to specific vibrations (e.g. 1120-1140 cm-1 can be C-C stretching) which molecules such as lipids may have, but using this as the basis for identifying a specific molecule seems to be an overinterpretation. Other fatty acids, for example, may have very similar peaks, within the error of the instrument especially.
-
Consistent with our multimodal analysis, senescent cells in wounded skin displayed increased Raman intensities at 1130/1134/1135 cm⁻¹, corresponding to 12-Methyltetradecanoic acid (15Aiso), 13-Methylmyristic acid (15iso), 15-Methylpalmitic acid (17iso),
How were these molecular identities obtained? I would be surprised if Raman intensities alone could support molecular assignments this specific.
-
We also observed Raman-only model in skin exhibited superior predictive performance relative to Raman-only model in lung across all evaluated metrics (e.g., 0.6000 vs. 0.4868 for Accuracy, 0.672 vs. 0.5367 for AUC). This enhanced predictive power for skin
Calling this "enhanced predictive power for skin" feels a bit misleading, as the AUC of 0.53 in lung corresponds to no predictive power.
-
using 70% of the cells for training and 30% for testing to compare the performance in classifying senescence cells
Does this include cells from all tissue and cell types? This seems like a major result, but the differences in performance are modest, so it would be informative to cross-validate on one of these dimensions (e.g. train on lung and predict for skin, and vice versa)
-
To evaluate the performance of the combined multimodal features in identifying senescent cells, we built a random forest classifier
I might be missing something, but it seems like this is a bit circular, as the features (DEGs and DPs) are selected using the same marker (p21+/-) that the model is predicting, so the feature space is "pre-aligned" to the prediction task.
-
The backscattered Raman light from the sample passes through two dichroic mirrors (DM1: Semrock LPD01 785RU 25, DM2: Semrock LPD01 785RU 25×36×1.1) and was collected by a multi-mode fiber (Thorlabs M14L 01). The collected signal was delivered to the imaging spectrograph (Holospec f/1.8i, Kaiser Optical Systems) and detected by a thermoelectric cooled, back illuminated and deep depleted CCD (PIXIS: 100BR_eXcelon, Princeton Instruments).
What is the spectral resolution and sampling rate of this system? The datasheet for this spectrograph lists a resolution of 3-6 cm^-1, and if the spectra have 873 dimensions and cover 600-1800 cm^-1, then the spectral sampling rate is around 1.4 cm^-1. Assuming these numbers are roughly correct, this makes it hard to interpret figures that highlight differences between Raman intensities at wavenumbers …
The backscattered Raman light from the sample passes through two dichroic mirrors (DM1: Semrock LPD01 785RU 25, DM2: Semrock LPD01 785RU 25×36×1.1) and was collected by a multi-mode fiber (Thorlabs M14L 01). The collected signal was delivered to the imaging spectrograph (Holospec f/1.8i, Kaiser Optical Systems) and detected by a thermoelectric cooled, back illuminated and deep depleted CCD (PIXIS: 100BR_eXcelon, Princeton Instruments).
What is the spectral resolution and sampling rate of this system? The datasheet for this spectrograph lists a resolution of 3-6 cm^-1, and if the spectra have 873 dimensions and cover 600-1800 cm^-1, then the spectral sampling rate is around 1.4 cm^-1. Assuming these numbers are roughly correct, this makes it hard to interpret figures that highlight differences between Raman intensities at wavenumbers closer than either of these values (e.g. 1643, 1644, and 1645 cm^-1 in Extended data Fig 11).
-
A total of 360 decreased and 422 increased DPs
Small comment: this is the first time the abbreviation "DP" is used, but what it stands for is not explained. Is it equivalent to the abbreviation "DPP" used in the methods section?
-
Together, these results indicate that aging in lung is generally associated with elevated Raman intensity at the tissue level but reduced intensity in specific immune population, whereas in skin, cell type specific spectral profiles highlight distinct metabolic adaptations and functional states that differentiate aging from senescence.
It would be helpful to explain the possible biological significance or interpretation of this difference, as it feels a little inscrutable at first glance.
-
This allowed us to compare Raman intensities between different cells at single-cell resolution.At the tissue level, old lung samples exhibited higher Raman intensity than young samples (46.7208±28.5472 vs. 43.2734±11.3974, p= 2.10E-13), whereas Raman intensity was lower in old T cells (42.4769±20.1176 vs. 43.0874±11.1290, p= 4.60E-07)
It seems like it would be important to normalize the mean Raman intensities by cell area or volume, otherwise changes in Raman intensity could be due to changes in cell area rather than a true change in the overall abundance of raman-active molecules or bonds.
-
