ISSN Print 2500–1094
ISSN Online 2542–1204
BIOMEDICAL JOURNAL OF PIROGOV UNIVERSITY (MOSCOW, RUSSIA)
ORIGINAL RESEARCH
Identification of microorganisms by Fourier-transform infrared spectroscopy
About authors
1 Emanuel Institute of Biochemical Physics of the Russian Academy of Sciences, Moscow, Russia
2 Bakulev National Medical Research Center of Cardiovascular Surgery, Moscow
3 Kulakov National Medical Research Center for Obstetrics, Gynecology and Perinatology, Moscow, Russia
4 Vavilov Institute of General Genetics, Russian Academy of Sciences, Moscow, Russia
5 Skryabin Moscow State Academy of Veterinary Medicine and Biotechnology, Moscow
Correspondence should be addressed: Alexei B. Shevelev
Kosygina 4, Moscow, 119334; moc.liamtoh@a_levehs
Received: 2018-08-01
Accepted: 2018-08-25
Published online: 2018-09-29
|
Fig. 1. The raw Fourier transform infrared spectra of pathogenic bacteria. Left to right and top to bottom are the spectra of A. baumannii, C. albicans, E. cloacae, E. coli, E. faecalis, E. faecium, K. pneumoniae, P. aeruginosa, S. aureus, S. epidermidis, and S. marcescens expressed in the units of light transmission
Fig. 2. The Fourier transform infrared spectra after preprocessing: normalization to the average transmission and calculation of the first derivative. Dotted lines show
the boundaries of the wavenumber ranges used in the analysis: 600, 1800, 2800, 3000 cm-1
Fig. 3. Projection of the initial data onto the linear discriminant (LD) obtained by PCA-LDA. Pathogens are marked by different colors. For the sake of convenience, the
spectra of the target pathogen S. aureus are separated from others on X-axis. The groups are well-separated given that LD > 0
Fig. 4. Linear discriminant coefficients obtained by PCA-LDA. These coefficients were used to obtain the projections in Fig. 3. The coefficients represent the informative
value of spectral data. For example, high values of the initial spectrum in the zone of negative coefficients indicate the probability that the studied isolate is not the
pathogen of interest
Fig. 5. Projection on the linear discriminant separating phenotypes MSSA and MRSA. Projections of MSSA spectra are shown in red; projections of MRSA spectra are shown in blue. Although 100% classification accuracy was not achieved, it was 80% given that LD > 0
