Abstract
Structural analysis of modified DNA with NMR is becoming ever more difficult with increasingly complex compounds under scrutiny for use in medical diagnosis, therapeutics, material science and chemical synthesis. To facilitate this process, we develop a molecular modeling approach to predict proton chemical shifts in sufficient agreement with experimental NMR measurements to guide structure elucidation. It relies on a QM/MM partitioning scheme and first principle calculations to predict the spatial structure and calculate corresponding proton chemical shifts. It is shown that molecular dynamics simulations that take into account solvent and temperature effects properly are of utmost importance to sample the conformational space sufficiently. The proposed computational procedure is universally applicable to modified oligonucleotides and DNA, attaining a mean error for the proton chemical shifts of less than 0.2 ppm. Here, it is applied on the Drew-Dickerson d(CGCGAATTCGCG)2 dodecamer as a benchmark system and the mispair-aligned N3T-ethyl-N3T cross-linked d(CGAAAT*TTTCG)2 undecamer, illustrating its universal use as computational tool to assist in structure elucidation. For the proton chemical shifts in the cross-linked system our methodology yields a strikingly superior description, surpassing the predictive power of (semi-)empirical methods. In addition, our methodology is the only one available to make an accurate prediction for the protons in the actual cross-link. To the best of our knowledge, this is the first computational study that attempts to determine the chemical shifts of oligonucleotides of this size and at this level of complexity.