Rapid Identification and Typing of Carbapenem-Resistant Klebsiella pneumoniae Using MALDI-TOF MS and Machine Learning

Abstract

Use matrix-assisted laser desorption ionisation time-of-flight mass spectrometry (MALDI-TOF MS) to screen the specific mass peaks of carbapenem-resistant Klebsiella pneumoniae (CRKP), compare the differences in spectrum peaks between intestinal and bloodstream screening of CRKP, and assess the utility of MALDI-TOF MS in quickly identifying various CRKP sources. From 2014 to 2023, a total of 267 Klebsiella pneumoniae strains were collected at Quzhou People's Hospital, including 60 intestinal screening isolates from ICU patients and 207 bloodstream infection isolates. MALDI-TOF MS was used to profile peptides in CRKP and carbapenem-sensitive Klebsiella pneumoniae (CSKP), followed by analysis with flexAnalysis and ClinProTools 3.0. Statistically significant protein peaks were selected to build classification models, which were verified using non-duplicate strains. MALDI-TOF MS achieved > 99.9% accuracy in identifying Klebsiella pneumoniae. Characteristic peaks (2523.43, 3041.62, 4520.11, 10,079.18 Da) were used to develop resistance analysis models, with the optimal model (SNN) showing 90.08% sensitivity, 95.80% specificity and identification accuracies of 90% for CSKP and 89.66% for CRKP. Another model using peaks (8876, 8993, 9139 Da) differentiated CRKP origins, with the ideal model (QC) achieving 86.85% sensitivity, 88.46% specificity, and accuracies of 81.82% for bloodstream and 95.00% for intestinal CRKP.

Keywords: Klebsiella pneumonia; GA; MALDI‐TOF MS; QC; SNN; bloodstream infection; carbapenemase; characteristic peak.

FIGURE 1

Distribution of bloodstream infection CRKP in clinical departments.

FIGURE 2

Antimicrobial susceptibility of isolates from different sources. (a) antimicrobial susceptibility of CRKP derived from bloodstream infection (b) from intestinal screening. ATM, aztreonam; CXM, cefuroxime; CZO, cefazolin; ETP, ertapenem; FOX, cefoxitin; GEN, gentamicin; MEM, meropenem; SXT, trimethoprim‐sulfamethoxazole; TOB, tobramycin; TZP, piperacillin/tazobactam.

FIGURE 3

MLST and MST of bloodstream infection CRKP.

FIGURE 4

Comparison of characteristic spectral peaks between CSKP and CRKP. (a) the ROC curve of the specific spectral peak at 5:2523.43 m/z (b) the ROC curve of peak 43: 4520.11 m/z (c) the contrast map of the specific spectral peak at 5:2523.43 m/z (d) the contrast map of peak 43:4520.11 m/z. The green line represents the mass spectrum characteristics of CSKP, and the red line represents the mass spectrum characteristics of CRKP.

FIGURE 5

Comparison of characteristics of CRKP from different sources. (a) the ROC curve of the specific spectral peak at 112:8993.44 m/z (b) the ROC curve of peak 115: 9139.06 m/z (c) The comparison of the spectrum peak at 108:8876.78 m/z and 112:8993.44 m/z (d) the comparison of the specific spectral peak at 115:9139.06 m/z (e) the two‐dimensional distribution of CRKP strains isolated from blood stream and intestinal tract based on ClinproTools analysis. The peaks at m/z 8993 and m/z 8877 served as the x‐ and y‐axes, respectively; Red and green indicated bloodstream infection CRKP and intestinal screening CRKP, respectively. The red line represents the mass spectrum characteristics of bloodstream infection CRKP, and the green line represents the mass spectrum characteristics of CRKP in intestinal screening.