| To demonstrate the utility of single-molecule fluorescence spectroscopy for the study of non-ideal samples and to overcome limitations of existing data-analysis techniques, measurements were made on two biologically relevant fluorophores at single molecule concentrations and new data-analysis techniques were developed.; In the analysis of DsRed, a new methodology was developed to measure dynamics within single-molecule events with μs resolution. It was found that at the single-molecule level, DsRed particles initially have a lifetime of about 3.6 ns and convert to a form having a lifetime of about 3.0 ns with a quantum yield of photoconversion on the order of 10−3 (calculated in terms of photons per DsRed tetramer). The particles then undergo additional photoconversion with a quantum yield of roughly 10−5 , generating a form with an average lifetime of 1.6 ns.; In the analysis of the intercalating homodimeric thiazole orange (TOTO) fluorophore with DNA at the single-molecule level, population analysis techniques were developed and used to quantify the relative binding affinity and properties upon binding to various DNA sequences. Unlike other analytical methods, this method is not limited by the small quantum yield of fluorescence that characterizes many biologically relevant fluorophores. By using measured distributions as a basis set, equilibrium binding affinities of TOTO were calculated. It was found that the fraction of poly-GC content in a molecule of dsDNA could be measured to within a few percent. It was also found that TOTO's binding affinity for certain 4-base pair sequences that had been shown previously to have extraordinarily high affinity for TOTO under conditions used in nuclear magnetic resonance experiments, was merely equal to that for poly-AT or poly-GC DNA under conditions more like those used in the biological labeling of nucleic acids.; The utility of neural-network-based analysis was also demonstrated for single-molecule event identification as compared to the maximum likelihood estimator (MLE) method. For ideal single-molecule data, neural networks and MLE were found to perform equally well. However, for non-ideal data, neural networks were able to identify single-molecule events more accurately than other data-analysis methods. |
