| The interaction between ethanol and the nicotinic acetylcholine receptor protein is detected either through NMR line broadening or T1 titration. Weighted Navon T1p analysis of T1 titration data indicates a dissociation constant consistent with published pharmacology, although Navon T1p analysis underestimates binding affinity. Nonlinear regression analysis of this data provides a best estimate of Kd = 55 muM at 4°C. within an unsymmetrical 90% confidence interval of [0.5, 440 muM].; T1 titration of the canon-binding subsite peptide, human neuromuscular junction alpha80--107, shows a maximum at which concentration 1H NOESY indicates a specific interaction with ethanol, as was seen for methanol. Nonlinear modeling of this data is consistent with a molecularity of two and with the known pharmacology. Nonlinear regression analysis provides a best estimate of Kd = 39 +/- 13 muM at 4°C, within the uncertainty range found for the nicotinic acetylcholine receptor.; Median estimation demonstrates that the error distribution of least squares T1 estimates is highly nongaussian. This observation led to the construction of spectral estimation (SE) of NMR relaxation, where in the domain of real error, optimal estimation based on prior knowledge is derived. Minimum bias and minimum variance are shown to be contradictory experimental design objectives. The analytical continuation of spectral estimation is constructed in an optimal manner, exploiting phase invariance. Under reasonable assumptions, spectral estimation is unbiased in the context of complex error and its variance is reduced because the modulus of the whole signal is used.; Simulations under difficult conditions validate and extend analytical predictions. These show that the imaginary Gaussian error distribution is commensurate with that of the relaxation rate (L). The estimation of amplitude (A), used to predict the estimation error from experimental data, is demonstrated analytically to be biased, in the context of real error. Furthermore, simulations demonstrate a 20% bias in the estimation of A in the context of complex Gaussian error when S/N is set to the low value of 10:1. Simulations were used to present the software concepts of DISPLAY and PICK, in a way that is more user friendly and efficient than NLS. |
