Beyond Amyloid: A Machine Learning-Driven Approach Reveals Properties of Potent GSK-3β Inhibitors Targeting...

Area of Science:

• Neuroscience
• Medicinal Chemistry
• Computational Biology

Background:

• Current Alzheimer's disease (AD) treatments only slow cognitive decline, lacking curative potential.
• Glycogen synthase kinase 3 beta (GSK-3β) is implicated in AD pathology through tau hyperphosphorylation and neurofibrillary tangles.
• Identifying inhibitors of GSK-3β is crucial for developing novel, non-amyloid-based AD therapeutics.

Purpose of the Study:

• To explore and curate common properties of active, drug-like molecules that inhibit GSK-3β.
• To utilize quantitative structure-activity relationship (QSAR) data and machine learning for predicting GSK-3β inhibitors.
• To identify key molecular features correlating with GSK-3β inhibitory activity.

Main Methods:

• Employed seven machine learning models: logistic regression, k-nearest neighbors, random forest, support vector machine, extreme gradient boosting, neural networks, and ensemble majority voting.
• Leveraged QSAR data from PubChem and ChEMBL databases.
• Performed feature importance analysis to identify critical molecular descriptors.

Main Results:

• The neural network (NN) model achieved the highest performance (79% AUC-ROC) on external validation data.
• Support vector machine (SVM) and random forest (RF) models showed superior compound classification and Kappa values.
• Feature analysis identified hydrogen bonds, phenol groups, and electronic characteristics as positively correlated with GSK-3β inhibition, while imidazole rings, sulfides, and methoxy groups showed negative correlation.

Conclusions:

• Structural, electronic, and physicochemical descriptors are significant for screening GSK-3β inhibitors.
• Predictive features identified can guide the development of novel therapeutic strategies targeting neurofibrillary tangles in AD.
• This research contributes to understanding essential properties for effective GSK-3β candidate inhibitors in non-amyloid-based AD treatments.