3D TOCSY-TOCSY processing using linear prediction, as a potential technique for automated assignment

Abstract

The assignment of the NMR spectra of proteins is the routine step in their structure determination (1, 2) and thus any approach to simplify or automate the evaluation of the spectra should be welcome. The first step in the assignment procedure is the proper recognition of the spin systems. This is commonly performed by 2D spectroscopy such as COSY, TOCSY, or relayed COSY (3-7) experiments and leads to a list of chemical shifts attributed to the protons involved in the coupling networks of the individual amino acids. There are already some approaches to automate this process (8-23), of which those based on “pattern recognition” in E.COSY (24) or z.COSY (25) spectra are the most prominent ones. The advantage of the latter procedure is the generality of the approach; i.e., any spin system may be recognized and the coupling network may be fully described. For proteins, this involves the recognition of the type of amino acid and the assignment of the resonances as NH, C,H, and CBH. Alternative approaches to the assignment of the spin systems are based on TOCSY and NOESY spectra which allow the collection of all chemical shifts of one residue, but may not always allow the correct assignment of the collected chemical shifts to the actual protons of the residue. Any approach based on 2D spectroscopy alone may suffer from the severe overlap of cross peaks usually observed in the aliphatic region of the spectra and may thus prevent the analysis of complicated coupling networks. Although the information contained in the relayed peaks of TOCSY or relayed spectra may help in overcoming ambiguities in some cases, the principal problem persists, especially as the peak-picking routines may no longer work due to overlap. For this reason, and because it is important to know the frequencies of the intermediate spins for the automated assignment, the consequent use of techniques yielding relayed peaks involves a three-dimensional Fourier transformation to resolve the overlap and to reveal the frequency of the intermediate spin (26-30). For the TOCSY technique, this can be achieved in a trivial manner by inserting an additional evolution time in the middle of the spin-lock period. Apart from the reduced amount of reasoning necessary by the computer there are a number of advantages which make the use of 3D TOCSY-TOCSY worthwhile: (i) peak-picking routines for 3D spectroscopy work more reliably, because lineshapes in three dimensions can be used as a criterion to recognize a peak; (ii) virtually no phase cycling is necessary and a three-dimensional spectrum with sufficient resolution can be recorded in a comparably short time; (iii)