Generation of an accurate CCSD(T)/CBS data set and assessment of DFT methods for the binding strengths of group I metal-nucleic acid complexes.

Accurate information about interactions between group I metals and nucleic acids is required to understand the roles these metals play in basic cellular functions, disease progression, and pharmaceuticals, as well as to aid the design of new energy storage materials and nucleic acid sensors that target metal contaminants, among other applications. From this perspective, this work generates a complete CCSD(T)/CBS data set of the binding energies for 64 complexes involving each group I metal (Li+, Na+, K+, Rb+, or Cs+) directly coordinated to various sites in each nucleic acid component (A, C, G, T, U, or dimethylphosphate). This data have otherwise been challenging to determine experimentally, with highly accurate information missing for many group I metal-nucleic acid combinations and no data available for the (charged) phosphate moiety. Subsequently, the performance of 61 DFT methods in combination with def2-TZVPP is tested against the newly generated CCSD(T)/CBS reference values. Detailed analysis of the results reveals that functional performance is dependent on the identity of the metal (with increased errors as group I is descended) and nucleic acid binding site (with larger errors for select purine coordination sites). Over all complexes considered, the best methods include the mPW2-PLYP double-hybrid and ωB97M-V RSH functionals (≤1.6% MPE; <1.0 kcal/mol MUE). If more computationally efficient approaches are required, the TPSS and revTPSS local meta-GGA functionals are reasonable alternatives (≤2.0% MPE; <1.0 kcal/mol MUE). Inclusion of counterpoise corrections to account for basis set superposition error only marginally improves the computed binding energies, suggesting that these corrections can be neglected with little loss in accuracy when using larger models that are necessary for describing biosystems and biomaterials. Overall, the most accurate functionals identified in this study will permit future works geared towards uncovering the impact of group I metals on the environment and human biology, designing new ways to selectively sense harmful metals, engineering modern biomaterials, and developing improved computational methods to more broadly study group I metal-nucleic acid interactions.

Assessment of Density Functional Theory Methods for the Structural Prediction of Transition and Post-Transition Metal-Nucleic Acid Complexes.

Understanding the structure of metal-nucleic acid systems is important for many applications such as the design of new pharmaceuticals, metal detection platforms, and nanomaterials. Herein, we explore the ability of 20 density functional theory (DFT) functionals to reproduce the crystal structure geometry of transition and post-transition metal-nucleic acid complexes identified in the Protein Data Bank and Cambridge Structural Database. The environmental extremes of the gas phase and implicit water were considered, and analysis focused on the global and inner coordination geometry, including the coordination distances. Although gas-phase calculations were unable to describe the structure of 12 out of the 53 complexes in our test set regardless of the DFT functional considered, accounting for the broader environment through implicit solvation or constraining the model truncation points to crystallographic coordinates generally afforded agreement with the experimental structure, suggesting that functional performance for these systems is likely due to the models rather than the methods. For the remaining 41 complexes, our results show that the reliability of functionals depends on the metal identity, with the magnitude of error varying across the periodic table. Furthermore, minimal changes in the geometries of these metal-nucleic acid complexes occur upon use of the Stuttgart-Dresden effective core potential and/or inclusion of an implicit water environment. The overall top three performing functionals are ωB97X-V, ωB97X-D3(BJ), and MN15, which reliably describe the structure of a broad range of metal-nucleic acid systems. Other suitable functionals include MN15-L, which is a cheaper alternative to MN15, and PBEh-3c, which is commonly used in QM/MM calculations of biomolecules. In fact, these five methods were the only functionals tested to reproduce the coordination sphere of Cu2+-containing complexes. For metal-nucleic acid systems that do not contain Cu2+, ωB97X and ωB97X-D are also suitable choices. These top-performing methods can be utilized in future investigations of diverse metal-nucleic acid complexes of relevance to biology and material science.

Assessment of the Accuracy of DFT-Predicted Li+-Nucleic Acid Binding Energies.

Understanding how lithium interacts with complex biosystems is crucial for uncovering the roles of this alkali metal in biology and designing extraction techniques for battery production and environmental remediation. In this light, fundamental information about Li+ binding to nucleic acids is required. Herein, a new database of Li+-nucleic acid interactions is presented that contains CCSD(T)/CBS benchmark energies for all nucleobase and phosphate binding locations. Furthermore, the performance of 54 DFT functionals in combination with three triple-zeta (TZ) basis sets (6-311+G(3df,2p), aug-cc-pVTZ, and def2-TZVPP) is tested. The results identify a range of functionals across different families (B2-PLYP, PBE-QIDH, ωB97, ωB97X-D, MN15, B3PW91, B97-2, TPSS, BP86-D3(BJ), and PBE) that can accurately describe coordinated Li+-nucleic acid interactions, with the average mean percent error (AMPE) across binding positions and basis sets being below 2%. Nevertheless, only three functionals tested (B2-PLYP, PBE-QIDH, and ωB97X-D) preserve this accuracy for metal cation-π interactions, suggesting that caution is warranted when choosing a functional to describe a diverse range of Li+-nucleic acid complexes. Removal of counterpoise corrections has very little impact on the reliability of most functionals, while the effect of empirical dispersion corrections varies depending on the functional choice and interaction type. While increasing the basis set to quadruple-zeta quality had little impact on the AMPE, the accuracy of double-zeta basis sets varies with family. Importantly, DFT methods reproduce the CCSD(T)/CBS trend in the preferred binding position for a given nucleic acid component and the global trend across components (phosphate ≫ G > C ≫ A ∼ T = U), as well as the geometries of the metal-nucleic acid complexes. The overall top performing functional is PBE-QIDH, which results in deviations from CCSD(T)/CBS values as small as ∼0.1 kcal/mol for nucleobase contacts and ∼1 kcal/mol for phosphate interactions. The most accurate DFT methods identified in the present work are recommended for future investigations of lithium interactions in larger nucleic acid systems to provide insights into the biological roles of this metal and the design of novel biosensing strategies.