Multidisciplinary characterization of potential inhibitors of the MSUT‐2 protein using molecular dynamics, mechanosensation assays, and deep AI‐based hit optimization for the treatment of Alzheimer’s disease

Abstract

AbstractBackgroundAlzheimer’s disease is a type of tauopathy, which is a broad category of diseases marked by abnormal aggregation and hyperphosphorylation of the tau protein. The MSUT‐2 protein has gained traction as a promising target for the treatment of tauopathies, given the protein’s RNA‐binding activity linked to tau aggregation. Previously in this research endeavor, using techniques like Quantum Machine Learning and Virtual Screening, 5 potential MSUT‐2 inhibitors have been identified from an initial set of 500,000 candidate compounds. The primary research aim is to further analyze these 5 potential hits using a new combination of computational methodologies and laboratory characterization.MethodA myriad of tests were performed with strains of C. elegans (microscopic roundworms that are established models for neurodegeneration) to validate previous computational indications. A series of motion evaluations, mechanosensation assays, and behavioral assessments were performed with CL2319 C. elegans (mutated to overexpress neuronal tau) with all 5 compounds. One of the compounds, named R128, was found to improve the health and behavior of tau mutant worms across experimental assessments at increasing doses of the compound. A 100 nanosecond simulation of the MSUT2‐R128 complex using the GROMOS96 43a1 force field and SPC Water Model was used to computationally validate protein‐ligand stability based on RMSD fluctuation values. DeepFrag, a fragment‐based lead optimization tool with a deep convolutional neural network backend, was then used to generate 20 potential optimized compounds, nine of which were found to retain R128’s druglikeliness profile yet have improved binding affinities to MSUT‐2.ResultThis multidisciplinary workflow has uncovered a promising hit for the MSUT‐2 protein, named R128, with 9 optimized derivatives with improved drug‐like qualities. This methodology can be used to efficiently identify promising treatment candidates for new protein targets involved in neurodegenerative diseases, with the prospect of expediting the drug discovery process significantly.ConclusionTo conclude, a translational approach, combining molecular dynamics, mechanosensation assessments with C. elegans, and an AI‐based hit optimization strategy has been used to identify new treatment candidates for Alzheimer’s disease. R128, a promising inhibitor for MSUT‐2 (an understudied target for Alzheimer’s disease), has also been identified in this study.