New Publication | Cell-Free SERS and Machine Learning Enable Bacterial Identification from Culture Supernatants

Website-ready release draft

We are excited to share our latest publication in Sensors and Actuators Reports, titled “Detection of Bacterial Signatures from Culture Supernatants Using Surface-Enhanced Raman Spectroscopy and Machine Learning.” 

Rapid and reliable bacterial identification is important for clinical diagnostics, food safety, veterinary medicine, and environmental monitoring. Traditional culture-based methods remain highly valuable but can require long processing times, while molecular and immunoassay methods often require target-specific reagents and sample preparation.

In this work, we demonstrate a simplified cell-free SERS strategy for bacterial identification. Instead of directly measuring intact bacterial cells, we analyze the culture supernatant—the liquid environment modified by bacterial growth through secreted metabolites, proteins, nucleic-acid fragments, phosphate-containing species, and other extracellular biomolecular products.

🔬What We Did

We collected cell-free culture supernatants from multiple bacterial groups, including foodborne pathogens E. coli O103, O111, O121, O157, and Salmonella STNR, as well as animal-associated Leptospira serovars Pomona, Bratislava, Canicola, Grippotyphosa, and Ictero. Matched sterile growth-media controls, including TSB and EMJH, were also analyzed.

The samples were diluted and deposited onto reproducible silver nanorod SERS substrates fabricated by oblique-angle deposition. The resulting SERS spectra were processed and analyzed using principal component analysis for clustering and support vector machine classification for bacterial identification.

⚡ Key Results

The culture supernatants produced reproducible and organism-specific SERS fingerprints. Compared with sterile media controls, the bacterial supernatants showed clear spectral changes, including peak shifts, intensity redistribution, and the emergence of phosphate-, nucleic-acid-, and protein-associated vibrational features.

Machine-learning analysis achieved strong classification performance:

• 99.29 ± 0.43% accuracy for foodborne bacterial supernatants

• 96.19 ± 1.79% accuracy for Leptospira serovar supernatants

• 97.83 ± 0.94% accuracy for the combined multi-group dataset acquired across independent laboratories

These results show that bacterial culture supernatants are not merely background signals in SERS measurements. Instead, they contain rich extracellular biochemical information that can serve as a reliable fingerprint for bacterial identification.

🔍Why This Matters

This study reframes bacterial culture supernatants as a useful analytical target for SERS-based biosensing. By measuring cell-free supernatants rather than intact cells, the workflow reduces direct handling of bacterial cells during the SERS measurement step while preserving diagnostically useful biochemical information.

More broadly, this work expands the practical scope of SERS and machine learning for bacterial detection. The approach may provide a foundation for future applications in food safety screening, veterinary diagnostics, environmental monitoring, and field-deployable biosensing platforms.

Read the full article: ScienceDirect article on Sensors and Actuators Reports.