There is currently a dearth of effective computational tools to design nucleobase-targeting small molecules and molecular mechanics force-fields for nucleobases lag behind their protein-focused counterparts. While quantum chemical methods can provide reliable interaction energies for small molecule-nucleobase interactions, these come at a steep computational cost. As a first step toward refining available tools for predicting small molecule-nucleobase interactions, we assessed the convergence of DFT-computed interaction energies with increasing binding site model size. We find that while accurate intercalator interaction energies can be derived from binding site models featuring only the flanking nucleotides for uncharged intercalators that bind parallel to the DNA base pairs, errors remain significant even when including distant nucleotides for intercalators that are charged, exhibit groove-binding tails that engage in non-covalent interactions with distant nucleotides, or that bind perpendicular to the DNA base pairs. Consequently, binding site models that include at least three adjacent nucleotides are required to consistently predict converged binding energies. The computationally inexpensive HF-3c method is shown to provide reliable interaction energies and can be routinely applied to such large models.<br>
On the Doubly hydrogen bonded dimer of 7-azaindole (0.1 M) as a model for DNA base pairs in acetonitrile solutions at rt
