Evaluation of Discrimination Performance in Case for Multiple Non-Discriminated Samples: Classification of Honeys by Fluorescent Fingerprinting

Elizaveta A Rukosueva,Valeria A Belikova,Andrei V Garmash,Ivan N Krylov,Mikhail K Beklemishev,Vladislav S Orekhov,Evgenii V Skorobogatov

doi:10.3390/s20185351

Abstract

In this study we develop a variant of fluorescent sensor array technique based on addition of fluorophores to samples. A correct choice of fluorophores is critical for the successful application of the technique, which calls for the necessity of comparing different discrimination protocols. We used 36 honey samples from different sources to which various fluorophores were added (tris-(2,2′-bipyridyl) dichlororuthenium(II) (Ru(bpy)32+), zinc(II) 8-hydroxyquinoline-5-sulfonate (8-Ox-Zn), and thiazole orange in the presence of two types of deoxyribonucleic acid). The fluorescence spectra were obtained within 400–600 nm and treated by principal component analysis (PCA). No fluorophore allowed for the discrimination of all samples. To evaluate the discrimination performance of fluorophores, we introduced crossing number (CrN) calculated as the number of mutual intersections of confidence ellipses in the PCA scores plots, and relative position (RP) characterized by the pairwise mutual location of group centers and their most distant points. CrN and RP parameters correlated with each other, with total sensitivity (TS) calculated by Mahalanobis distances, and with the overall rating based on all metrics, with coefficients of correlation over 0.7. Most of the considered parameters gave the first place in the discrimination performance to Ru(bpy)32+ fluorophore.

Highlights

Fluorescence spectrum of a sample comprises a specific fingerprint, which can be used for classification purposes
We considered the number of such intersections as a measure of discrimination performance
Intrinsic fluorescence of 36 honey samples did not allow for their complete discrimination using

Summary

Introduction

Fluorescence spectrum of a sample comprises a specific fingerprint, which can be used for classification purposes. Most fluorimetric fingerprinting classification methods use intrinsic fluorescence of samples [1,2]. A less developed approach is based on adding fluorophores to samples [3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22], which promises their better discrimination because of the interaction of fluorophore(s) with non-fluorescent components of the sample and corresponding changes in the spectrum. A key issue in the development of the “add-a-fluorophore” strategy is selection of fluorophores, since their chemical nature determines the possibility of fluorescence quenching/dequenching by the sample components.

Methods

Discussion

Conclusion