top of page
Search

New Publication | Label-Free SERS and Machine Learning Enable Rapid BRDC Virus Classification and Quantification

Bovine respiratory disease complex, or BRDC, is one of the most important infectious disease challenges in cattle health. It is a multifactorial disease involving environmental stress, host immune response, viral infection, and secondary bacterial infection. Rapid and pathogen-specific identification of BRDC-associated viruses is important for veterinary diagnostics, cattle health management, and future point-of-care disease monitoring.

In this work, we demonstrate a label-free SERS and machine-learning platform for direct detection, classification, and concentration prediction of major BRDC-associated viruses. Instead of relying on antibodies, reporter labels, or target-specific reagents, the method uses intrinsic SERS spectral fingerprints from viral specimens measured on silica-coated silver nanorod substrates.


What We Did

We analyzed five major BRDC-associated viruses:

  • Bovine respiratory syncytial virus (BRSV)

  • Bovine viral diarrhea virus type 1 (BVDV-1)

  • Bovine viral diarrhea virus type 2 (BVDV-2)

  • Infectious bovine rhinotracheitis virus (IBRV)

  • Bovine parainfluenza virus type 3 (BPIV-3)

Serially diluted virus specimens were deposited onto AgNR@SiO₂ SERS substrates fabricated by oblique angle deposition. The silica coating was used to improve substrate stability during virus incubation. SERS spectra were collected using a portable 785 nm Raman system with a 1 s acquisition time per spectrum.

The spectra were processed using consistent preprocessing steps, including baseline correction, interpolation, and normalization. Machine-learning models were then developed for two complementary tasks: SVM-based virus classification and SVR-based concentration prediction.


⚡ Key Results

The five BRDC viruses produced reproducible and virus-specific SERS fingerprints, even though they share many common biochemical components such as proteins, lipids, and nucleic acids.

Machine-learning analysis achieved strong performance:

  • 99.57% overall SVM classification accuracy

  • No misclassification between viral spectra and background controls

  • R² = 0.9974 for SVR-based viral concentration prediction

  • MAE = 0.028 and RMSE = 0.0373 for concentration regression

  • 100% blind specimen-level virus identification using majority-vote SVM analysis

  • Successful concentration-trend prediction for withheld dilution levels not included in model training

These results show that label-free SERS spectra contain sufficient molecular information to distinguish closely related BRDC-associated viruses and to support quantitative concentration prediction over multiple dilution levels.


Why This Matters

This study provides a proof-of-concept foundation for rapid, label-free, and machine-learning-assisted veterinary virus diagnostics. Compared with conventional molecular assays, the SERS approach does not require amplification, fluorescent probes, antibodies, or target-specific labels during spectral measurement.

More broadly, this work extends our SERS and AI diagnostic framework from human respiratory viruses to major veterinary pathogens. The approach may support future development of field-compatible or pen-side diagnostic tools for cattle health monitoring, outbreak response, and precision livestock disease management.


Read the full article: Amit Kumar, Shaun Steven Van Den Hurk, Fengbo Ma, Yanjun Yang, Xianyan Chen, Binu T Velayudhan, Hemant K. Naikare, Ralph A. Tripp, Yiping Zhao, “Label-Free Surface-Enhanced Raman Spectroscopy and Machine Learning for Rapid Classification and Quantification of Bovine Respiratory Disease Complex Viruses,” Talanta, (2026)



 
 
 

Comments

bottom of page