Class identification of biomolecules based on multicolor native fluorescence spectroscopy
All biological molecules are composed of only a few basic building blocks and, therefore, exhibit similar physical properties. In particular, aromatic amino acids can be found in biological molecules and exhibit native fluorescence. Tryptophan, phenylalanine and tyrosine are some of the most common fluorescent molecules. In addition many enzymes or cofactors, NADH (reduced b-nicotinamide adenine dinucleotide) and Riboflavin being the most prominent examples, exhibit pronounced native fluorescence. Laser induced fluorescence (LIF) is a promising tool for differentiating fluorescing molecules. However, the variety of biological molecules is huge compared to the number of basic building blocks. Therefore, the fluorescence spectra of different analytes are often quite similar, and high spectral resolution of the emission spectra is required to reveal differences. Additionally multi-wavelength LIF is essential to improve bio-agent class identification. In order to test class identification based on multi-color native fluorescence we have measured, with a conventional laboratory set-up, the laser-induced fluorescence spectra of various biological building blocks (e.g., tryptophan, tyrosine, flavins, NADH, riboflavin, and chloroform). We have recorded fluorescence spectra for representative bio-agent simulants in solution for various classes of biomolecules (i.e., toxins, viruses, bacteria, and fungi). As simulants we have used, for example, bovine serum albumin (BSA), horse heart cytochrome C (HCC), bacillus thuringiensis (simulant for anthrax) and yeast. We have begun creating a library of fluorescence spectra that were recorded for different excitation wavelengths, and we have identified distinguishing features in the spectra. In addition, differences and correlations of these spectra have been evaluated by using statistical methods (e.g., the principal component analysis (PCA)). Based on theses measurements and evaluations we have defined a set of characteristic features suitable for class discrimination. In contrast to other approaches which rely mostly on spectral filters and record only limited spectral information, we have collected detailed spectral information for each excitation wavelength. This allows us to define more specific principle components (e.g., certain fluorescence intensities of different spectral regions within one spectrum or between spectra from different excitation wavelength). As shown in Fig. 1, the PCA evaluation leads to clear separation between the simulants even when we have used only the native fluorescence spectra obtained with 266nm excitation. By evaluating the fluorescence spectra obtained with both 266nm and 355nm excitation, even better discrimination between the simulants has been achieved. Future issues to be addressed include randomized testing on unknown analytes in order to explore the identification probability, recording fluorescence spectra for analyte mixtures, evaluating the class identification probability for mixtures of analytes, and examining class identification for simulants in different solutions (e.g., di-ionized and tap water).
Bassler, M. ; Schmidt, O. ; Kiesel, P. ; Johnson, N. M. Class identification of biomolecules based on multicolor native fluorescence spectroscopy. ISSSR 2006 (International Symposium on Spectral Sensing Research); 2006 May 29-June 2; Bar Harbor; ME; USA.