Simultaneous CEA and CDH-17 Detection from Cancer Extracellular Vesicles Via Machine Learning Enabled Bio Silent Azide-Infrared Spectroscopy
7 affiliations
- 1Department of Chemistry, Shiv Nadar Institution of Eminence, Uttar Pradesh 201314, Delhi-NCR, India
- 2Department of Chemistry
- 3Indian Institute of Technology Bhilai
- 4Department of Bioscience and Biomedical Engineering
- 5All India Institute of Medical Sciences, Department of Biophysics, India
- 6Department of Electrical and Computer Engineering
- 7University of Cyprus
Abstract
Extracellular vesicles (EVs) provide a rich, minimally invasive source of cancer biomarkers, yet EV surface glycoproteins remain comparatively underexploited compared with EV internal cargo. Here, we report a bio silent, azide-based Fourier transform infrared (FTIR) spectroscopic platform for simultaneous, label-minimal detection of two colorectal cancer (CRC) EV surface biomarkers, carcinoembryonic antigen (CEA) and cadherin-17 (CDH-17), directly from intact vesicles. EVs from five CRC cell lines and a non-cancerous control line were first distinguished by their intrinsic biochemical fingerprints across the lipid, protein, carbohydrate, and nucleic acid spectral regions, achieving up to 97% classification accuracy by multivariate (PCA/SIMCA) and machine-learning (PLS-DA) analysis. Selective azide labeling of CEA and CDH-17 via biotinylated antibodies and streptavidin–azide generated a specific vibrational signal in the 2000–2300 cm⁻¹ bio-silent window without perturbing native EV spectra; peak authenticity was confirmed by Gaussian deconvolution, quantitative signal-to-noise analysis (SNR = 315 ± 99), and buffer/serum-artifact controls. Density-of-states overlap, cosine correlation, and constrained non-negative least-squares regression enabled simultaneous deconvolution and quantification of both biomarkers from mixed-antibody spectra, resolving cell line–specific expression patterns and enabling EV subtyping. Critically, the platform was validated on plasma-derived EVs from CRC patients and a healthy donor, where diagnostic performance exceeded that of the cell-line panel (AUC = 1.0 for metastatic CRC vs. healthy across all six spectral ratios; AUC = 0.58–1.0 for early-stage CRC vs. healthy). These results establish azide-FTIR spectroscopy as a rapid, label-efficient, and artifact-controlled platform for multiplexed EV biomarker detection, with direct evidence of its translatability to clinical liquid biopsy specimens rather than relying on cell-line data alone.
Related
- Single molecule detection of Tau seeds in extracellular vesicles from Alzheimer's disease brains
- Machine-learning-enhanced Optical Photothermal infrared (O-PTIR) spectroscopy for rapid simultaneous identification of multiple metabolites to monitor biotransformations: proof of concept on Brewer's spent grain (BSG)
- Bimodal spectroscopy integrating multi-wavelength time-resolved photoacoustic spectroscopy and near-infrared spectroscopy with deep learning for quantitative detection of serum biochemical indicators
- Editorial: Investigating the role of extracellular vesicles in cardiovascular disease mechanisms
- ICAM-5: A marker for neuronally derived extracellular vesicles
BibTeX
@misc{paul2026simultaneous,
title = {{Simultaneous CEA and CDH-17 Detection from Cancer Extracellular Vesicles Via Machine Learning Enabled Bio Silent Azide-Infrared Spectroscopy}},
author = {Debashish Paul and Natasha Natasha and Sunita Saha and Suchetan Pal and Ananya Ananya and Saroj Kumar and Chrysafis Andreou and Tatini Rakshit},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15000620/v2},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15000620/v2},
note = {CC BY},
}
RIS
TY - UNPB
TI - Simultaneous CEA and CDH-17 Detection from Cancer Extracellular Vesicles Via Machine Learning Enabled Bio Silent Azide-Infrared Spectroscopy
AU - Debashish Paul
AU - Natasha Natasha
AU - Sunita Saha
AU - Suchetan Pal
AU - Ananya Ananya
AU - Saroj Kumar
AU - Chrysafis Andreou
AU - Tatini Rakshit
PY - 2026
DA - 2026/09/30
PB - ChemRxiv
DO - 10.26434/chemrxiv.15000620/v2
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15000620/v2
AB - Extracellular vesicles (EVs) provide a rich, minimally invasive source of cancer biomarkers, yet EV surface glycoproteins remain comparatively underexploited compared with EV internal cargo. Here, we report a bio silent, azide-based Fourier transform infrared (FTIR) spectroscopic platform for simultaneous, label-minimal detection of two colorectal cancer (CRC) EV surface biomarkers, carcinoembryonic antigen (CEA) and cadherin-17 (CDH-17), directly from intact vesicles. EVs from five CRC cell lines and a non-cancerous control line were first distinguished by their intrinsic biochemical fingerprints across the lipid, protein, carbohydrate, and nucleic acid spectral regions, achieving up to 97% classification accuracy by multivariate (PCA/SIMCA) and machine-learning (PLS-DA) analysis. Selective azide labeling of CEA and CDH-17 via biotinylated antibodies and streptavidin–azide generated a specific vibrational signal in the 2000–2300 cm⁻¹ bio-silent window without perturbing native EV spectra; peak authenticity was confirmed by Gaussian deconvolution, quantitative signal-to-noise analysis (SNR = 315 ± 99), and buffer/serum-artifact controls. Density-of-states overlap, cosine correlation, and constrained non-negative least-squares regression enabled simultaneous deconvolution and quantification of both biomarkers from mixed-antibody spectra, resolving cell line–specific expression patterns and enabling EV subtyping. Critically, the platform was validated on plasma-derived EVs from CRC patients and a healthy donor, where diagnostic performance exceeded that of the cell-line panel (AUC = 1.0 for metastatic CRC vs. healthy across all six spectral ratios; AUC = 0.58–1.0 for early-stage CRC vs. healthy). These results establish azide-FTIR spectroscopy as a rapid, label-efficient, and artifact-controlled platform for multiplexed EV biomarker detection, with direct evidence of its translatability to clinical liquid biopsy specimens rather than relying on cell-line data alone.
ER -