Rapid discrimination of Pseudomonas aeruginosa ST175 isolates involved
in a nosocomial outbreak using MALDI-TOF Mass Spectrometry and FTIR
Spectroscopy coupled with Machine Learning
Objectives: Evaluation of Matrix-Assisted Laser Desorption
Ionization-Time of Flight Mass Spectrometry (MALDI-TOF MS) and Fourier
Transform Infrared-Spectroscopy (FTIR-S) as diagnostic alternatives to
DNA-based methods for the detection of Pseudomonas aeruginosa
sequence type (ST) 175 isolates involved in a hospital outbreak.
Methods: Twenty-seven P. aeruginosa isolates from a 2014
outbreak in the Hematology department of our hospital were previously
characterized by PFGE and WGS. Besides, 8 P. aeruginosa isolates
were analyzed as unrelated controls. MALDI-TOF MS spectra were acquired
by applying the colony on the MALDI target plate followed by 1 µl of
formic acid 100% and 1 µl of HCCA matrix. For the analysis with FTIR-S,
colonies were resuspended in 70% ethanol and sterile water according to
the manufacturer instructions. Spectra from both methodologies were
analyzed using Clover Biosoft® software, that allowed data modelling
using different algorithms and validation of the classifying models.
Results: Three outbreak-specific biomarkers were found at 5169,
6915 and 7236 m/z in MALDI-TOF MS spectra. Classification models
based on these three biomarkers showed the same discrimination power
displayed by PFGE. Besides, K-Nearest Neighbor algorithm allowed the
discrimination of the same clusters provided by whole-genome sequencing
and the validation of this model achieved 97.0% correct classification.
On the other hand, FTIR-S showed a discrimination power similar to PFGE
and reached correct discrimination of the different STs analyzed.
Conclusions: The combination of both technologies evaluated,
paired with Machine Learning tools, may represent a powerful tool for
real-time monitoring of high-risk clones and isolates involved in