Rapid Classification of Multilocus Sequence Subtype for Group B Streptococcus Based on MALDI-TOF Mass Spectrometry and Statistical Models

Abstract

Group B Streptococcus (GBS) is an important etiological agent of maternal and neonatal infections as well as postpartum women and individuals with impaired immunity. We developed and evaluated a rapid classification method for sequence types (STs) of GBS based on statistic models with Matrix-Assisted Laser Desorption/Ionization Time-of Flight Mass Spectrometry (MALDI-TOF/MS). Whole-cell lysates MALDI-TOF/MS analysis was performed on 235 well-characterized GBS isolates from neonatal invasive infections in a multi-center study in China between 2015 and 2017. Mass spectra belonging to major STs (ST10, ST12, ST17, ST19, ST23) were selected for model generation and validation. Recognition and cross validation values were calculated by Genetic Algorithm-K Nearest Neighbor (GA-KNN), Supervised Neural Network (SNN), QuickClassifier (QC) to select models with the best performance for validation of diagnostic efficiency. Informative peaks were further screened through peak statistical analysis, ST subtyping MSP peak data and mass spectrum visualization. For major STs, the ML models generated by GA-KNN algorithms attained highest cross validation values in comparison to SNN and QC algorithms. GA-KNN models of ST10, ST17, and ST12/ST19 had good diagnostic efficiency, with high sensitivity (95-100%), specificity (91.46%-99.23%), accuracy (92.79-99.29%), positive prediction value (PPV, 80%-92.68%), negative prediction value (NPV, 94.32%-99.23%). Peak markers were firstly identified for ST10 (m/z 6250, 3125, 6891) and ST17 strains (m/z 2956, 5912, 7735, 5218). Statistical models for rapid GBS ST subtyping using MALDI-TOF/MS spectrometry contributes to easier epidemical molecular monitoring of GBS infection diseases.

Keywords: MALDI-TOF/MS; group B Streptococcus; multi-locus sequence typing; sequence type; statistical model.

Figure 1

Gel overview and peak statistic of the GBS MALDI-TOF mass spectra for different ST classes of GBS strains. (A) Gel overview (B) Peak statistic overview of the 235 mass spectra for six ST classes of GBS strains. ST10 (dark green), ST12 (purple), ST17 (yellow), ST19 (blue), ST23 (light green), other STs (red). (C) Peak statistic overview of the spectra for ST10 model, ST10 (green), non-ST10(red); (D) ST12 model, ST12 (green), non-ST12 (red); (E) ST17 model, ST17 (green), non-ST17(red); (F) ST19 models, ST19 (green), non-ST19 (red); (G) ST12/ST19 model, ST12/ST19 (green), non-ST12/ST19 (red). The other STs group included 39 GBS isolates belong to 26 other ST types ST2, ST4, ST8, ST24, ST27 , ST55, ST88, ST103, ST138, ST146, ST156, ST163, ST179, ST188, ST197 , ST223, ST249, ST268, ST335, ST357, SST452, ST480, ST579, ST651, ST680, ST938.

Figure 2

Spectra plots showing the presence or absence of the relevant peak biomarkers for MALDI-TOF/MS discrimination of the four main GBS clonal complexes in the genetic algorithm model. ST10 (red), ST12 (green), ST17 (blue); ST19 (yellow). x-axis shows the mass per charge ratio values (m/z) and y-axis indicates the intensities of peaks expressed in arbitrary intensity units. (A) Peak biomarkers for ST10 (m/z 3125, 6250, 6891) and ST12 (m/z 4510); (B) Peak biomarkers for ST17 (m/z 2956, 5218, 5912, 7735); (C) Peak biomarkers for ST10 and ST12 (m/z 9017), ST17 (m/z 7620).

Figure 3

Comparison of the peak biomarkers used to detect ST1, ST10, ST12, ST17, ST19 of GBS strains by MALDI-TOF MS (m/z ±10) ^aRed, specific presence of the peak; green, specific absence of the peak; blue, peak included in the typing model; orange, overexpressed peak; purple, under-expressed peak.