structures for: Towards biochemically relevant QM computations on nucleic acids. Controlled electronic structure geometry optimizations of nucleic acids structural motifs using penalty restraint functions.
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Recent developments of dispersion-corrected density functional theory methods allow for the first time to describe large fragments of nucleic acids (hundreds of atoms) with an accuracy clearly surpassing the accuracy of common biomolecular force fields. Such calculations can significantly improve the description of the potential energy surface of nucleic acids molecules, which may be useful for studies of molecular interactions and conformational preferences of nucleic acids, as well as verification and parameterization of other methods. The first of such studies, however, demonstrated that successful applications of accurate QM calculations to larger nucleic acids building blocks are hampered by difficulties to obtain geometries that are biochemically relevant and are not biased by non-native structural features. We present an approach that can greatly facilitate large-scale QM studies on nucleic acids, namely, electronic structure geometry optimizations of nucleic acid fragments utilizing a penalty function to restrain key internal coordinates with a specific focus on the torsional backbone angles. The work explores the viability of these restraint optimizations for DFT-D3, PM6-D3H and HF-3c optimizations on a set of examples (an UpA dinucleotide, a DNA G-quadruplex and a B-DNA fragment). Evaluation of different penalty function strengths reveals only a minor system-dependency and reasonable restraint values range from 0.01 to 0.05 Eh/rad2 for the backbone torsions. Restraints are crucial to perform the QM calculations on biochemically relevant conformations in implicit solvation and gas phase geometry optimizations. The reasons why to use restrained instead of constrained or unconstrained optimizations are explained and an open-source external optimizer is provided.