Thomas Wenzel, Bruker Daltonik GmbH, Fahrenheitstrasse 4, 28359 Bremen, Germany
Gongyi Shi, Bruker Daltonics Inc., 2859 Bayview Dr., Fremont, CA 94538
Stefan Klepel, Bruker Daltonik GmbH, Fahrenheitstrasse 4, 28359 Bremen, Germany
Simone Stumpf, Bruker Daltonik GmbH, Fahrenheitstrasse 4, 28359 Bremen, Germany
Thomas Maier, Bruker Daltonik GmbH, Fahrenheitstrasse 4, 28359 Bremen, Germany
Markus Kostrzewa, Bruker Daltonik GmbH, Fahrenheitstrasse 4, 28359 Bremen, Germany |
|
MALDI-TOF mass spectrometry fingerprint analysis for the identification of microorganisms recently has emerged as a powerful tool in clinical microbiology diagnostics. Starting with a single colony from a culture plate and using minimal sample preparation, characteristic profile spectra are acquired for an unknown microbe and identification is performed by bioinformatic comparison with the entries of a dedicated database. One current drawback of the technology is that in the case of mixed cultures, e.g. contaminations of picked colonies, generally only one of the microorganisms is detected in automated analyses, sometimes even any identification is hampered by the mix of profiles in the mass spectra. On the other hand, careful evaluation of the mass spectra derived from two or even three different bacteria in a mixture can be unravelled after inspection by eye. Here, we present a bioinformatic approach to check automatically spectra for the probable existence of such contaminations. Defined mixtures of bacteria were prepared in a relation of 1:1 up to 10:1 and subsequently analysed using a Microflex benchtop mass spectrometer according to standard procedures. The resulting spectra were automatically analysed with the MALDI Biotyper 2.0 software package and, in parallel, with a novel algorithm which is calculating the possibility of a mixed-culture spectrum. Further, spectra derived from pure cultures were analysed with the novel control algorithm to evaluate the risk of “false alarms”. After parameter optimization, known mixed-culture spectra derived from routine samples, detected by careful visual analysis earlier, were investigated with the improved mixture detection algorithm. The results show the general applicability of the approach to real-life samples if the parameters for the mixture detection are chosen carefully.
|
|
