Sensible Functional Linear Discriminant Analysis Effectively Discriminates Enhanced Raman Spectra of Mycobacterium Species

Abstract

Tuberculosis caused by Mycobacterium tuberculosis complex (MTBC) is one of the major infectious diseases in the world. Identification of MTBC and differential diagnosis of nontuberculous mycobacteria (NTM) species impose challenges because of their taxonomic similarity. This study describes a differential diagnosis method using the surface-enhanced Raman scattering (SERS) measurement of molecules released by Mycobacterium species. Conventional principal component analysis and linear discriminant analysis methods successfully separated the acquired spectrum of MTBC from those of NTM species but failed to distinguish between the spectra of different NTM species. A novel sensible functional linear discriminant analysis (SLDA), projecting the averaged spectrum of a bacterial specie to the subspace orthogonal to the within-species random variation, thereby eliminating its influence in applying linear discriminant analysis, was employed to effectively discriminate not only MTBC but also species of NTM. The successful demonstration of this SERS–SLDA method opens up new opportunities for the rapid differentiation of Mycobacterium species.