None of the current force fields are perfect, and many challenges persist in DNA and RNA simulations. Broadly, the inaccuracies can be attributed to two main sources: dihedral space parameterization and the parameterization of intermolecular interactions. Here we focus on dihedral space parameterization only, but it is important to note that several well-documented artifacts in force fields stem from inaccuracies in intermolecular potentials.
DNA
The canonical B-DNA double helix is reasonably well described by many existing force fields, even those that perform poorly for non-canonical DNA structural motifs. The OL force field variants have not only helped to improve the description of canonical B-DNA, but provide a much more reliable representation of non-canonical DNA structures, such as guanine quadruplexes, Z-DNA and non-canonical backbones observed in protein-DNA complexes. While the torsion space has been significantly improved by OL24 (all backbone torsions have been modified compared to ff99), some problems remain in describing intermolecular interactions.
RNA
The conformational space of RNA is highly complex, and many challenges persist. Only a few of the existing loop structures are described correctly and many non-canonical motifs, such as GpU platform in sarcin/ricin loop (SRL) do not behave well in MD simulations. Given the numerous inaccuracies observed in well studied small RNA motifs, it is unlikely that unfolded RNA can be accurately modeled. Despite intense parameterization efforts, capturing the fine structural details of RNA continues to be a significant challenge.
Feedback on the performance of our force fields and especially pointing us to new difficult cases is greatly appreciated.
