Evaluation of Fourier-Transform Infrared Spectroscopy with IR Biotyper® system for Streptococcus pneumoniae serotyping.

Purpose: Streptococcus pneumoniae is an important human pathogen and its virulence is centered on the expression of a polysaccharide capsule. Serotyping pneumococci has clinical and epidemiological relevance. As the gold standard (Quellung reaction) is expensive and somehow laborious, alternative methodologies for timely serotyping are desirable. We aimed to evaluate IR Biotyper^® (Bruker), which is based on FT-IRS, as a tool for serotyping S. pneumoniae by customizing a classifier focusing on PCV15 serotypes.

Methods: Initially, we evaluated the Pneumococci serogrouper (Bruker) included in the IR Biotyper^® software (v.4.0) using 188 retrospective isolates of S. pneumoniae previously serotyped by the gold standard. Lately, these isolates were used to construct (n = 83) and validate (n = 105) the PCV15-Classifier. Overall, 1246 spectra were obtained to create a classification model using artificial neural network (ANN) machine learning in 400 cycles. Finally, prospective S. pneumoniae (n = 61), recovered from patients attended at hospitals in South Brazil, were used for a "real-life" evaluation of both classifiers.

Results: Agreement was higher when using PCV15-Classifier instead of Pneumococci serogrouper (Bruker) among both, retrospective (86.7% versus 68.1%; p = 0.0004) and prospective isolates (75.4% versus 70.5%; p = 0.54), highlighting (i) its better performance for epidemiologically relevant serotypes, such as 3, and (ii) its capacity to discriminate between serotypes of serogroup 19 (19F and 19A).

Conclusion: Altogether, our results reiterate the usefulness of IR Biotyper^® for serotyping S. pneumoniae applying a customized classifier, which can be re-trained whenever needed, following the evolution of serotype distribution.