Nonfitting protein–ligand interaction scoring function based on first‐principles theoretical chemistry methods: Development and application on kinase inhibitors

Abstract

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.

Parametrization of an Orbital-Based Linear-Scaling Quantum Force Field for Noncovalent Interactions

“…The autodock program 98 was used to dock 73 ligands to the protein structure using autodock’s default options. The ligands were chosen as a subset of those listed in ref (41) subject to the availability of DFTB3 parameters. One of these ligands, ligand 29 in ref (41) is chosen to be a “lead”, i.e., a ligand that has been identified as well-inhibiting the function of the protein.…”

“…The ligands were chosen as a subset of those listed in ref (41) subject to the availability of DFTB3 parameters. One of these ligands, ligand 29 in ref (41) is chosen to be a “lead”, i.e., a ligand that has been identified as well-inhibiting the function of the protein. The goal then, is to identify additional candidate ligands that are structurally similar to the lead ligand but are predicted to inhibit the protein function with greater efficacy.…”

“…28−31 Others have used LS-QM methods to examine interactions within clusters, 32−36 compute chemical reaction energies, 37−40 and perform drug screening. 41−43 …”

Benchmark of electronic structure methods for protein–ligand interactions based on high-level reference data

“…4,5 Many commonly used \scoring functions" perform reasonably well for this task, but no single approach seems to be generally applicable when high accuracy is needed. 6,7 Especially the last¯ve years have seen a number of studies on how to improve upon these methods with computationally more demanding quantum chemistry based approaches, [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] but many questions are still open in this¯eld. We have recently published a large-scale study using systematically generated model systems from the PDBbind database 23,24 and comparing molecular mechanics (MM), semiempirical quantum mechanical (SQM) and density functional theory (DFT) methods.…”

The amyloid interactome: Exploring protein aggregation

“…The amyloid interactome displays the interacting partners of in vivo amyloid forming proteins in a flat and detailed protein map (Fig 1). The results presented in this work, combine interactions from specialized networks of protein aggregation [34, 35, 37–40, 64] and eventually, assemble a new set of functionally unconnected proteins into a network that would possibly fill the missing pieces of protein aggregation and shed light towards the exploitation of novel disease protein-targets.…”