Nonfitting protein-ligand interaction scoring function based on first-principles theoretical chemistry methods: development and application on kinase inhibitors.

Targeted therapy is currently a hot topic in the fields of cancer research and drug design. An important requirement for this approach is the development of potent and selective inhibitors for the identified target protein. However, current ways to estimate inhibitor efficacy rely on empirical protein-ligand interaction scoring functions which, suffering from their heavy parameterizations, often lead to a low accuracy. In this work, we develop a nonfitting scoring function, which consists of three terms: (1) gas-phase protein-ligand binding enthalpy obtained by the eXtended ONIOM hybrid method based on an integration of density functional theory (DFT) methods (XYG3 and ωB97X-D) and the semiempirical PM6 method, (2) solvation free energy based on DFT-SMD solvation model, and (3) entropy effect estimated by using DFT frequency analysis. The new scoring function is tested on a cyclin-dependent kinase 2 (CDK2) inhibitor database including 76 CDK2 protein inhibitors and a p21-activated kinase 1 (PAK1) inhibitor database including 20 organometallic PAK1 protein inhibitors. From the results, good correlations are found between the calculated scores and the experimental inhibitor efficacies with the square of correlation coefficient R(2) of 0.76-0.88. This suggests a good predictive power of this scoring function. To the best of our knowledge, this is the first high level theory-based nonfitting scoring function with such a good level of performance. This scoring function is recommended to be used in the final screening of lead structure derivatives.