Frequency-domain resolution of protein fluorescence

Multi-exponential decays of fluorescence can also be recovered by measurements in the frequency-domain. This has only become practical within the past four years [27,28]. The resolution of multi-exponential decays requires measurements over a wide range of light modulation frequencies. Earher instrumentation could operate at only one to three frequencies, and these hmited data were not adequate to determine the four parameters in equation 10 (two and two t,). The new instruments 

Typical frequency-domain data for two proteins are shown in Fig. 17. The data consist of the phase angles and modulation, each measured over the widest possible range of frequencies. This requirement illustrates the transform relationship between the time and the frequency-domain measurements. In the time-domain, the most desirable excitation profile is the shortest obtainable pulse. The Fourier transform of a 5-function consists of all frequencies. Hence, the experimental requirements are similar, short pulses or wide range frequencies. For each protein (Fig. 17) the phase angle increases and the modulation decreases as the frequency increases. The data are analyzed in a manner analogous to the time-domain data. That is, a decay law is assumed and the parameters varied until the best possible match is obtained 

Both melittin and S, Nuclease contain a single tryptophan residue. The data illustrated the point that the emission from such simple proteins can be multiexponential. Even though only a single residue is responsible for the emission, it was not possible to fit the data using a single decay time. This is shown by the failure of 

It is important to exercise clear thinking when fluorescence data in either domain are fitted to various models. A poor fit can be used to reject a model. The poor fit can be due to either an inadequate model or due to systematic errors in the data not known to the researcher. If systematic errors occur a more complex model could be accepted to account for the errors, not because the model is appropriate for the sample. Secondly, a good fit does not prove the model which yields the good fit is correct. A good fit only shows that the model is adequate to explain the data. Alternatively stated, the data which yield the good fit are not adequate to support a more complex model. 

---