AI-Powered SERS Spectroscopy Breakthrough Boosts Safety of Medicinal Food Products

AI-powered SERS boosts safety of medicinal food products © Udomner-chronicles-stock.adobe.com

A cross-institutional research team has unveiled a powerful new technology that could dramatically improve the safety and authenticity of medicinal and edible homologs (MEHs)—natural products widely used in both cuisine and traditional medicine. Combining surface-enhanced Raman spectroscopy (SERS) with artificial intelligence, the new “SERSome” system delivers fast, accurate identification of MEHs, a crucial advancement in preventing the accidental use of toxic look-alikes and counterfeit ingredients (1).

Published in Talanta, the study was led by Shuang Jiang, Yue Zhao, Qingyu Meng, Rongheng Ma, Huimin Yu, and Yang Li et al. The work was conducted at The Second Affiliated Hospital of Harbin Medical University, Zhejiang University School of Medicine, Harbin Medical University, and the University of Oulu in Finland (1).

SERSome Represents a New Era for Molecular Fingerprinting
At the core of the research is the SERSome platform, which leverages the molecular fingerprinting capability of surface-enhanced Raman spectroscopy (SERS) to analyze MEHs. Traditional SERS methods have been hampered by signal interference and inconsistent results, especially when dealing with complex mixtures like those found in MEHs. To overcome these challenges, the researchers created a stable and interference-resistant SERS substrate by synthesizing silver nanoparticles using sodium borohydride as both reducer and activator. The addition of calcium ions induced nanoparticle aggregation, creating “hotspots” that dramatically enhanced signal strength (1).

These enhancements allowed the system to avoid common issues such as fluorescence interference and inconsistent signal amplification, resulting in a more reproducible and sensitive detection method. The team reported a detection limit as low as 100 femtograms per milliliter, a significant milestone for trace analysis (1).

Deep Learning Meets Spectroscopy
To process the vast amount of spectroscopic data generated by SERS, the researchers integrated a deep learning algorithm into the platform. The resulting AI-powered system could accurately identify 77 different MEHs in a newly created SERS spectral database. Utilizing a K-BPNN (K = 7) model and t-SNE for dimensionality reduction, the system achieved an identification accuracy of 98% when paired with minimal manual oversight (1).

This deep learning layer enabled label-free, automated identification that circumvents the subjectivity and inefficiency of traditional methods. As a result, MEHs such as Astragali radix, Angelicae sinensis radix, and Lycii fructus—which often share visual similarities with toxic species—can now be reliably differentiated (1).

Addressing a Public Health Need
MEHs are increasingly used worldwide for their therapeutic properties, from reducing inflammation to supporting post-COVID recovery. However, their safety has been compromised by the proliferation of visually similar toxic plants and counterfeit products. Notable examples include confusion between Phytolaccae radix and Panax ginseng, or the toxic Abrus precatorius seeds mistaken for the benign Vignae semen (1).

The SERSome platform provides an urgently needed tool for real-time, on-site authentication of MEHs, protecting consumers and enabling regulatory agencies to conduct efficient quality control (1).

Limitations and Future Development
Despite its impressive accuracy, the researchers acknowledge current limitations. The system has only been tested on 77 MEHs, a small fraction of the total in circulation. As the spectral library grows, maintaining the same accuracy will require further refinement of both the deep learning models and the SERS substrate. Still, the team is optimistic, citing ongoing efforts to scale the system for broader applications in medicinal plant quality assurance (1).
By integrating advanced nanomaterials chemistry with cutting-edge artificial intelligence (AI), the SERSome platform offers a groundbreaking approach to MEH identification and safety assurance. As the MEH market expands globally, this technology could play a key role in safeguarding public health and enabling the full potential of medicine-food homology to be realized (1). AI-powered Raman Spectroscopy in the form of SERS is becoming a powerful tool for analysis in clinical and health-related situations (1,2)

References

(1) Jiang, S.; Zhao, Y.; Meng, Q.; Ma, R.; Sun, X.; Lyu, X.; Zhang, L.; Wu, G.; Wang, X.; He, Y.; Liang, J. Advanced SERSome-Based Artificial-Intelligence Technology for Identifying Medicinal and Edible Homologs. Talanta 2025, 292, 127931. https://doi.org/10.1016/j.talanta.2025.127931.

(2) Workman, J., Jr. New AI-Powered Raman Spectroscopy Method Enables Rapid Drug Detection in Blood. Spectroscopy Online, February 10, 2025. https://www.spectroscopyonline.com/view/new-ai-powered-raman-spectroscopy-method-enables-rapid-drug-detection-in-blood (accessed 2025-04-09).