ML188 for COVID-19

The indispensable role of the SARS-CoV-2 main protease (Mpro) in the viral replication cycle and its dissimilarity to human proteases make Mpro a promising drug target. In order to identify the non-covalent Mpro inhibitors, we performed a comprehensive study using a combined computational strategy. We first screened the ZINC purchasable compound database using the pharmacophore model generated from the reference crystal structure of Mpro complexed with the inhibitor ML188. The hit compounds were then filtered by molecular docking and predicted parameters of drug-likeness and pharmacokinetics. The final molecular dynamics (MD) simulations identified three effective candidate inhibitors (ECIs) capable of maintaining binding within the substrate-binding cavity of Mpro. We further performed comparative analyses of the reference and effective complexes in terms of dynamics, thermodynamics, binding free energy (BFE), and interaction energies and modes. The results reveal that, when compared to the inter-molecular electrostatic forces/interactions, the inter-molecular van der Waals (vdW) forces/interactions are far more important in maintaining the association and determining the high affinity. Given the un-favorable effects of the inter-molecular electrostatic interactions—association destabilization by the competitive hydrogen bond (HB) interactions and the reduced binding affinity arising from the un-compensable increase in the electrostatic desolvation penalty—we suggest that enhancing the inter-molecular vdW interactions while avoiding introducing the deeply buried HBs may be a promising strategy in future inhibitor optimization.

Abstract Due to the lack of a method to efficiently represent the multimodal information of a protein, including its structure and sequence information, predicting compound-protein binding affinity (CPA) still suffers from low accuracy when applying machine-learning methods. To overcome this limitation, in a novel end-to-end architecture (named FeatNN), we develop a coevolutionary strategy to jointly represent the structure and sequence features of proteins and ultimately optimize the mathematical models for predicting CPA. Furthermore, from the perspective of data-driven approach, we proposed a rational method that can utilize both high- and low-quality databases to optimize the accuracy and generalization ability of FeatNN in CPA prediction tasks. Notably, we visually interpret the feature interaction process between sequence and structure in the rationally designed architecture. As a result, FeatNN considerably outperforms the state-of-the-art (SOTA) baseline in virtual drug evaluation tasks, indicating the feasibility of this approach for practical use. FeatNN provides an outstanding method for higher CPA prediction accuracy and better generalization ability by efficiently representing multimodal information of proteins via a coevolutionary strategy.