In this communication we describe a new scheme to process the data from stimulated echo protein diffusion experiments. For a series of gradient-encoded proton spectra \(\:{f}_{k}\left(\omega\:\right)\) considered over the selected spectral region \(\:({\omega\:}_{left},{\omega\:}_{right})\) , we build a model to approximate the unique (protein-dependent) shape of the spectrum. Taking a cue from the optimal filtration theory, \(\:{f}_{model}\left(\omega\:\right)\) is constructed as the intensity-weighted combination of \(\:{f}_{k}\left(\omega\:\right)\) . The so obtained \(\:{f}_{model}\left(\omega\:\right)\) is then used to fit the individual spectra \(\:{f}_{k}\left(\omega\:\right)\) , thus providing highly accurate estimates for the integral signal intensities that are subsequently used for Stejskal-Tanner-type analyses. This algorithm has been implemented as a part of a new web server, named DDfit (https://ddfit.org, mirror at https://ddfit.bio-nmr.spbu.ru/). The server accepts spectrometer data from the standard stimulated and double-stimulated echo experiments by Bruker, as well as custom-designed experiments. The server is easy to use, with data processing taking no more than several seconds. Our tests using simulated as well as experimental data found that DDfit determines protein diffusion coefficients with both accuracy and precision, offering several-fold improvement in precision compared to other processing schemes.
