Rapid discrimination of Klebsiella pneumoniae carbapenemase 2-producing and non-producing Klebsiella pneumoniae strains using near-infrared spectroscopy (NIRS) and multivariate analysis.

Klebsiella pneumoniae Carbapenemase (KPC-2)-producing and non-producing Klebsiella pneumoniae (KP) have rapidly disseminated worldwide, challenging the diagnostics of Gram-negative infections. We evaluate the potential of a novel non-destructive and rapid method based on Near-Infrared Spectroscopic (NIRS) and multivariate analysis for distinguishing KPC-2-producing and non-producing KP. Thirty-nine NIRS spectra (24 KPC-2-producing KP, 15 KPC-2 non-producing KP) were acquired; different pre-processing methods such as baseline correction, derivative and Savitzky-Golay smoothing were performed. A spectral region fingerprint was achieved after using genetic algorithm-linear discriminant analysis (GA-LDA) and successive projection algorithm (SPA-LDA) algorithms for variable selection. The variables selected were then used for discriminating the microorganisms.Accuracy test results including sensitivity and specificity were determined. Sensitivity in KPC-2 producing and non-producing KP categories was 66.7% and 75%, respectively, using a SPA-LDA model with 66 wavenumbers. The resulting GA-LDA model successfully classified both microorganisms with respect to their "fingerprints" using only 39 wavelengths. Sensitivity in KPC-2 producing category was moderate(≈66.7%) using a GA-LDA model. However, sensitivity in KPC-2 non-producing category using GA-LDA accurately predicted the correct class (with 100% accuracy). As100% accuracy was achieved, this novel approach identifies potential biochemical markers that may have a relation with microbial functional roles and means of rapid identification of KPC-2 producing and non-producing KP strains.