A Combination of MALDI-TOF MS Proteomics and Species-Unique Biomarkers’ Discovery for Rapid Screening of Brucellosis

Abstract

Brucellosis is considered to be a zoonotic infection with a predominant incidence in most parts of Iran that may even simply involve diagnostic laboratory personnel. In the present study, we apply matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF MS) for rapid and reliable discrimination of Brucella abortus and Brucella melitensis, based on proteomic mass patterns from chemically treated whole-cell analyses. Biomarkers of the low molecular weight proteome in the MALDI-TOF MS spectra were assigned to conserved ribosomal and structural protein families that were found in genome assemblies of B. abortus and B. melitensis in the NCBI database. Significant protein mass signals successfully mapped to ribosomal proteins and structural proteins, such as integration host factor subunit alpha, cold-shock proteins, HU family DNA-binding protein, ATP synthase subunit C, and GNAT family N-acetyltransferase, with specific biomarker peaks that have been identified for each virulent and vaccine strain. Web-accessible bioinformatics algorithms, with a robust data analysis workflow, followed by ribosomal and structural protein mapping, significantly enhanced the reliable assignment of key proteins and accurate identification of Brucella species. Furthermore, clinical samples were analyzed to confirm the most dominant protein biomarker candidates and their relevance for the identifications of B. melitensis and B. abortus. With proper optimization, we envision that the presented MALDI-TOF MS proteomics analyses, coupled with special usage of bioinformatics, could be used as a cost-efficient strategy for the diagnostics of brucellosis and introduce a reliable identification protocol for species of dangerous bacteria.

Human brucellosis and fever of unknown origin

“…Given the atypical symptoms of brucellosis, it is difficult for clinicians to distinguish a group of diseases with similar symptoms. Although in recent years Hamidi et al [ 11 ] and Kazemi et al [ 12 ] showed that the discovery of some potential markers can play an effective role in the rapid diagnosis of brucellosis. However, at present, these methods are rarely actually used in clinical practice, especially in non-endemic areas of brucellosis.…”

Rapid, automatic typing ofClostridioides difficileRibotypes Using MALDI-TOF MS

“…The first was Clover MSDAS, a commercial ad hoc software developed by Clover Bioanalytical Software (Granada, Spain, https://cloverbiosoft.com) for pre-processing of the protein spectra, biomarker detection and automatic typing with well-known ML algorithms: Random Forest (RF), K-Nearest Neighbor (KNN), Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machine (SVM), and Light Gradient Boosting Machine (Light-GBM). In Clover MSDAS, all algorithms underwent training using biomarkers identified by the software, as described in previous studies (Hamidi, Bagheri Nejad et al 2022; Busby, Doyle et al 2023; Candela, Arroyo et al 2023). The use of these biomarkers for differentiation of specific categories is referred to as Expert-Knowledge (EK).…”

“…In Clover MSDAS, all algorithms underwent training using biomarkers identified by the software, as described in previous studies (Hamidi, Bagheri Nejad et al 2022;Busby, Doyle et al 2023;Candela, Arroyo et al 2023). The use of these biomarkers for differentiation of specific categories is referred to as Expert-Knowledge (EK (auto, sqrt, log2).…”

Algorithms push forward the application of MALDI–TOF mass fingerprinting in rapid precise diagnosis

“…78 It can be achieved by pretreatment of the specimens using strategies like differential centrifugation, affinity capture, magnetic precipitation, microchip separation, in-situ extraction, chromatography, and so on. [79][80][81][82][83] A large number of patient specimens are also required to establish reliable machine learning classifiers with sufficient enough learning data.…”

“…The success of such a biomarker discovery methodology relies heavily on the samples pretreatment procedures to obtain as many mass peaks as possible, especially the peaks from low‐abundant proteins and peptides 78 . It can be achieved by pretreatment of the specimens using strategies like differential centrifugation, affinity capture, magnetic precipitation, microchip separation, in‐situ extraction, chromatography, and so on 79–83 . A large number of patient specimens are also required to establish reliable machine learning classifiers with sufficient enough learning data.…”