New Methods For Diffusion Ordered NMR Spectroscopy

NMR has been widely applied to a variety of fields,such as Physics,Chemistry,Biology,Material and Environment science.As a noninvasive technique,NMR not only quantifies and quantitates compounds in mixtures,but also plays an important and unique role in studying the structure and dynamics of biomolecules at the atomic level.Diffusion-ordered NMR spectroscopy(DOSY)is a powerful tool to investigate a mixture,and to identify the peaks of one chemical component without physical separation like optical spectroscopy techniques.The effectiveness of DOSY spectrum depends strongly on the processing method used,especially in the case of spectral overlap that is unavoidable for mixture.There are two kinds of methods that have been developed to overcome the problem of spectral overlapping:experimental design and data post-processing.Data post-processing methods do not take any extra experimental time and,therefore,attracting more and more attention.Multivariate methods make use of spectral difference between components to resolve overlapped peaks.However,for all the multivariate based post-processing methods,the maximum number of resolvable components is limited to 2-5 in practice.Dividing the spectrum into many individual segments and then analyzing each individually before combining the results is an effective strategy.But how to accurately segment the spectra and determine the number of components(NC)of the segment remains ambiguous.Accurate estimate of NC makes the segmentation much easier and greatly facilitates the subsequent data processing.In general,many data mining methods are capable of determining NC.But these methods often underestimate NC in case of more than 2 or 3 components because the diffusion decay vectors are severely collinear.According to our tests of the simulated and experimental datasets.Principle component analysis can be used to give the minimal value of NC.Joint approximate diagonalization of eigenmatrices(JADE)is best among them but is sensitive to experimental error.Principal component analysis is able to estimate the minimum number of components.To make full use of the difference between component spectra and the difference of diffusion coefficients,We proposed a novel post-processing method,simultaneous inversion of Laplace transform(SILT).SILT is evaluated on both simulated and experimental datasets.It is demonstrated that SILT can handle the overlapped spectra with up to 6 components in spite of few artifacts.In addition to its ability to more accurately estimate NC,the new method can generate ID spectrum for each of the overlapped components.Furthermore,SILT can handle both discrete and continuous diffusion coefficients and has much higher resolution than the inversion of Laplace transform,provided that there is obvious spectral difference between components.Thus,SILT is less insensitive to the deviation from the exponential model.Last,the performance of DOSY is subject to many factors,an important one of which is chemical exchange.Chemical exchange is a common phenomenon in nature,such as protein-ligand association and dissociation processes.Determining chemical exchange rate facilitates to elucidate the mechanism of molecular interactions.Accurate quantification of the chemical exchange rate is difficult because relaxation is always involved in the diffusion process.Herein,Genetic Algorithm(GA)was introduced to determine the exchange rate constants.The diffusion coefficient without chemical exchange is determined by experimental technique.It was found that the values of chemical exchange rate constants were largely underestimated using DOSY dataset;compared with Levenberg Marquardt(LM)method that does not converge sometimes,the GA method we introduced is more stable and accurate.