M. OBERLE¹, N. WOHLWEND², D. JONAS³, F. MAURER⁴, G. JOST⁵, K. VRANCKX⁶, A. EGLI<sup>7

1</sup>Medizinische Mikrobiologie, Institut für Labormedizin, Kantonsspital Aarau, Aarau, Schweiz, ²Labormedizinisches Zentrum Dr. Risch AG, Liebefeld-Bern, Schweiz, ³Institute for Environmental Medicine, University of Freiburg, Germany, ⁴Institute for Medical Microbiology, University of Zurich, Switzerland, ⁵Clinical Microbiology, Dianalabs, Geneva, Switzerland, ⁶Applied Maths NV, Sint-Martens-Latem, Belgium, ⁷Clinical Microbiology, University Hospital Basel, Basel, Switzerland

Due to the wide access to MALDI-TOF, an increasing number of studies show reliable typing results. This ‘new’ method holds promise for an easy access typing method compared to conventional approaches. However, the technical, biological and center reproducibility of such typing data has not yet been explored. The aim of this study is to compare typing data from a multi-center study employing bioinformatics.
Seven different center, tested a set of 12 extended spectrum beta-lactamase (ESBL)-producing Escherichia coli isolates – including two outbreak and one non-related cluster. Every center used a Microflex MALDI-TOF system (Bruker, Bremen, Germany) and was instructed to use a previously evaluated standard operating procedure. Each center recorded four spectra from each isolate (technical replicate) and repeated the procedure at three different days (biological replicate). The raw data was used to calculate the technical and biological reproducibility in each center and between the centers using the software Bionumerics (Applied Maths, Belgium).
The median values of the technical reproducibility ranged between 96.8 % and 99.4% for each center. The median values of the biological reproducibility ranged between 47.6 % and 94.4 %. The clustering among identical isolates was reproducible between centers, however a more complex bioinformatics approach was necessary:
We used a discriminant analysis to better separate the three clusters. For this, we defined a reference center who’s spectra were used to identify 18 discriminating peaks using an Anova analysis. With this approach, the identification of the cluster was 100% correct.
Detection of specific clusters is possible from spectra obtained from different centers. However, bioinformatic techniques are required to make the analysis robust and reliable. The closer related different clusters are to each other, the less reliable their separation is. Technical reproducibility is a key element to implement MALDI-TOF based typing. Further examinations are needed to study which elements during the preparation and the acquirement of spectra contribute most to the technical variability.