Our new paper has just been published. We investigate the correlation between the Voronoi entropy (VE) of ligand molecules and their affinity to receptors to test the hypothesis that less ordered ligands have higher mobility of molecular groups and therefore a higher probability of attaching to receptors. VE of 1144 ligands from a database of potential anti-Covid drugs is calculated using SMILES-based 2D graphs representing the molecular structure. The affinity of the ligands with the SARS-CoV-2 main protease is obtained from the BindingDB Database as half-maximal inhibitory concentration (IC50) data. The VE distribution is close to the Gaussian, 0.4 ≤ Sv ≤ 1.66, and a strong correlation with IC50 is found, IC50 = −275 Sv + 613 nM, indicating the correlation between ligand complexity and affinity. On the contrary, the Shannon entropy (SE) descriptor failed to provide enough evidence to reject the null hypothesis (p-value > 0.05), indicating that the spatial arrangement of atoms is crucial for molecular mobility and binding.
In a sense, ligands are treated in this work as a text build of polygons as alphabetic elements.
S. Shityakov, A. S. Aglikov, E. V. Skorb, and Michael Nosonovsky* 2023 “Voronoi Entropy as a Ligand Molecular Descriptor of Protein–Ligand Interactions” ACS Omega, https://doi.org/10.1021/acsomega.3c07328
https://pubs.acs.org/doi/full/10.1021/acsomega.3c07328
There is a significant correlation between the Voronoi Entropy and affinity to the SARS-CoV-2 main protease (Mpro) represented by IC50.
Docking of the compounds with the Mpro protein. Protein–ligand binding site is shown as black mesh
No significant correlation with Shannon Entropy (SE) of the SMILES language description of ligands in the database of 1144 molecules. This suggests that the linear chain of characters in the SMILES language, while containing information about molecules, is insufficient to predict their binding properties. One needs a spatial arrangement (which is captured by the Voronoi Entropy but not by the SMILES formula).