Fourier-transform infrared spectroscopy with machine learning for rapid O-serogrouping and typhi differentiation of...

Salmonella enterica remains a leading cause of gastrointestinal infections worldwide, requiring rapid and accurate typing for effective outbreak management and clinical treatment. Traditional serological methods, such as the Kauffmann-White scheme, are labor-intensive, time-consuming, and require specialized expertise. This study evaluated the IR Biotyper (IRBT) system, which combines Fourier-transform infrared (FTIR) spectroscopy with machine learning algorithms, for rapid classification of clinical Salmonella isolates at the O-serogroup level and for differentiating S. Typhi from non-typhoidal Salmonella (NTS) in a routine pediatric microbiology setting in the Middle East. A total of 164 clinical isolates from blood and stool cultures were analyzed, generating 1,925 spectra for serogroup classification and 790 spectra for the Typhi vs. NTS classifier. Using the Wellcolex™ Color Salmonella kit as a reference, the IRBT O-group classifier achieved an overall accuracy of 98.2%, correctly identifying 161 of 164 isolates. The system reached 100% accuracy for Groups A (n = 4, 95% CI: 39.8%-100%), A+E/G (n = 5, 95% CI: 47.8%-100%), B (n = 40, 95% CI: 91.2%-100%), C+D (n = 13, 95% CI: 75.3%-100%), D (n = 49, 95% CI: 92.7%-100%), and isolates expressing the Vi antigen (n = 2, 95% CI: 15.8%-100%), while Group E/G showed lower accuracy at 89.5%. The Typhi vs. NTS classifier achieved 100% accuracy in both internal validation (n = 34, 95% CI: 89.8%-100%) and independent testing (n = 31, 95% CI: 88.8%-100%), demonstrating complete concordance with the Vi antiserum assay. These results indicate that FTIR spectroscopy offers a rapid, reliable, and user-friendly alternative to conventional serotyping, reducing turnaround time to under 2 h. Integration of this technology into clinical microbiology workflows provides a robust tool for real-time pathogen monitoring and enhanced public health response.