Computational prediction of vibrational spectra of normal and modified DNA nucleobases

Purines, pyrimidines, and the corresponding ribose monophosphates are ubiquitous biomolecules involved in several cellular processes–in DNA and RNA, signaling, and energy transactions. In the new and exciting field of DNA‐inspired nanostructures, they are used as fundamental building blocks. Unique features of the nucleobases are that several tautomeric states are close in energy and the tautomeric equilibria are sensitive to exocyclic substitution and pH. Knowing the exact structure and tautomer(s) at physiological conditions is crucial to understand the substrate specificities and catalytic mechanisms of the many enzymes for which these nucleobases are substrates. Very few spectroscopic methods can distinguish between tautomers and provide solution structures. Vibrational spectroscopy has long been known to be an excellent tool to obtain reliable information on nucleic acids. However, even when good‐quality spectra are available, isotope editing is required to make reliable band assignments and identify structures unequivocally. Density functional theoretical (DFT) methods have become indispensible in assisting the assignment of observed spectra to normal modes, identification of tautomers, and modeling of spectra of isotope‐edited molecules. We review the performance of DFT methods in the prediction of nucleobases and their analogs. We find that even with modest basis sets, trends in vibrational spectra can be predicted adequately and guide assignments to normal modes of the molecule. Shifts in band positions induced upon isotope labeling are reproduced more reliably than the band positions. Scaling considerably improves the agreement between the computed and experimental spectra, but accurate prediction of vibrations of exocyclic groups like CO requires that solvent effects are taken into account. Copyright © 2009 John Wiley & Sons, Ltd.