Computational identification of inhibitors of MSUT‐2 using quantum machine learning and molecular docking for the treatment of Alzheimer's disease

Abstract

AbstractBackgroundAlzheimer's Disease is in a class of related neurodegenerative diseases known as tauopathies, where tau hyperphosphorylation leads to progressive neurological decline. MSUT‐2 has emerged as a viable target for the treatment of tauopathies like Alzheimer's, as the protein's interactions with the tau protein could be central to the accumulation and aggregation of tau. Quantum Computing and Machine Learning are two developing fields that have the capability to completely revolutionize the field of pharmacology. Developed architectures like Quantum Support Vector Machines can be scaled up readily as technology improves, further enhancing this breakthrough field’s implications on pharmacology.MethodIn the first phase of research, OpenEye's FILTER software was employed, which eliminated a set of starting compounds from ChemBridge's EXPRESS‐PICK Screening Library (over 500,000 compounds) that were predicted to be highly toxic and have low bioavailability. Next, a simple pharmacophore screen was performed using a set of known partial actives and inactives on OpenEye's ROCS tool. A modified Quantum Support Vector Machine algorithm was then run through IBM Qiskit. The model was trained using dozens of known chemical features of FDA‐approved chemical actives and inactives. Compounds were featurized using data from SwissADME. Following preprocessing and training, the model was executed to predict known actives from the remaining set of test compounds. Finally, compounds were subject to a precise molecular docking trial with MSUT‐2 (using OpenEye’s FRED tool and AUTODOCK Vina), where binding interaction strength was directly evaluated.ResultThis computational screening methodology resulted in numerous predicted inhibitors of MSUT‐2. To validate the methodology of this screening workflow, numerous known FDA‐approved drugs and actives for other diseases were successfully identified using the exact screening methodology utilized in this experiment through retrospective analysis.ConclusionIn conclusion, a promising computational screening methodology partially validated with translational data using advancements in Quantum Machine Learning has been identified. This exact workflow is novel and promising, and can easily be scaled up for comprehensive screening for a myriad of targets. 5 inhibitors for a promising yet underexplored target for Alzheimer’s disease, MSUT‐2, have also been identified in this study.