In nuclear magnetic resonance (NMR) studies of proteins, the assignment of nuclear spin resonances to individual atoms is a prerequisite for all subsequent biomedical applications, such as studies of ligand binding, protein-DNA interactions, and dynamics. Resonance assignment is one of the most time consuming and labor intensive steps even when the 3D xray structure of the protein is available. A new strategy is proposed to find resonance assignments that makes optimal use of the x-ray structure, residual dipolar coupling measurements, and chemical shifts. In this way, the complementary strengths of NMR, x-ray crystallography, and quantum chemistry are synergetically used. In contrast to standard assignment protocols, no NMR information about sequential connectivities is required. The assignment problem is mathematically formulated in terms of a weighted matching problem that can be solved using a computationally efficient combinatorial optimization algorithm. The chemical shift information provided by the assignment can be directly used for a wide variety of NMR applications including ligand binding studies that help to characterize binding sites, strengths, and specificity that will help to rationally guide drug design. The proposed work will make a large number of proteins, whose structure is deposited in the protein database (PDB), amenable to detailed biomedical NMR investigations. Structural genomics, which aims at the derivation of protein function from its 3D structure, requires rapid structure determination methods. For proteins whose structure is not determined by x-ray crystallography, an efficient method is proposed for the simultaneous structure determination and resonance assignment using NMR residual dipolar coupling information, incomplete sequential connnectivities, and modeling techniques. The method avoids the slow assignment step by directly building a library of 3D protein fragments that are consistent with residual dipolar couplings, amino-acid type specific chemical shifts, and sparse sequential backbone connectivity information. The individual fragments have good resolution and they are not biased towards known structures deposited in the PDB. The protein fragments are then assembled to complete 3D protein structures by high-performance computional methods using database derived penalty functions. In contrast to standard NMR methods, this approach does not rely on NOESY-derived distance constraints. The method, which promises a significant speed-up over standard NMR methods, will be tested and refined on model proteins and then applied to biologically relevant proteins.
Showing the most recent 10 out of 53 publications
