Potential bias in NMR relaxation data introduced by peak intensity analysis and curve fitting methods

We present an evaluation of the accuracy and precision of relaxation rates calculated using a variety of methods, applied to data sets obtained for se veral very different protein systems. We show that common methods of data e valuation, such as the determination of peak heights and peak volumes, may be subject to bias, giving incorrect values for quantities such as R-1 and R-2. For example, one common method of peak-height determination, using a s earch routine to obtain the peak-height maximum in successive spectra, may be a source of significant systematic error in the relaxation rate. The alt ernative use of peak volumes or of a fixed coordinate position for the peak height in successive spectra gives more accurate results, particularly in cases where the signal/noise is low, but these methods have inherent proble ms of their own. For example, volumes are difficult to quantitate for overl apped peaks. We show that with any method of sampling the peak intensity, t he choice of a 2- or 3-parameter equation to fit the exponential relaxation decay curves can dramatically affect both the accuracy and precision of th e calculated relaxation rates. In general, a 2-parameter fit of relaxation decay curves is preferable. However, for very low intensity peaks a 3 param eter fit may be more appropriate.