An Innovative Inter-Spectral Distance-Based Approach to Analyzing MALDI-TOF Mass Spectra of Closely Related Microbial...

Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is a highly efficient technique for microbial identification; however, the accuracy has always been a problem when identifying closely related microbial species. Improving spectral data identification algorithms is one of the key approaches to enhancing the discriminatory power and reliability of identification for the closely related species.

This study develops a dimensionality reduction method based on inter-spectral distance computation for the analysis of MALDI-TOF MS data. The method comprises four steps: average spectrum construction, peak matching, distance calculation, and spectral vectorization. We applied this method, along with the conventional principal component analysis (PCA) method, to a MALDI-TOF MS dataset of closely related microbial species. Binary classification experiments were conducted to compare the classification performance of the two methods, and multiclass classification experiments were conducted to evaluate the feasibility of the proposed approach for database construction.

A systematic evaluation of the newly proposed distance-based method was conducted using MALDI-TOF mass spectral data from five pairs of closely related microbial species. The results indicated that this method effectively extracted spectral features and enabled accurate classification. It outperformed the conventional PCA method, and even other more sophisticated methods like LDA and t-SNE, in terms of both clustering performance and identification accuracy.

The findings suggest that the newly proposed distance-based dimensionality reduction algorithm (DbDRA) largely enhances the reliability of identifying closely related microbial species, highlighting its potential applicability in microbial identification using MALDI-TOF mass spectroscopy.