|
Designing novel GSK-3β inhibitors based on artificial intelligence and molecular modeling
|
Mohammad Azimi1   , Ahmad Ebadi2 |
1- Hamadan University of Medical Sciences
2- Hamadan University of Medical Sciences , ahmadebadie@gmail.com |
|
|
Abstract: (9 Views) |
| In this study, an innovative Artificial Intelligence-based framework was employed to design and evaluate novel GSK-3β enzyme inhibitors with potential for Parkinson’s disease treatment. Initially, a set of new compounds was generated de novo using chemical language models (CLMs). These compounds were then refined and evaluated based on drug-likeness criteria, structural diversity, and physicochemical properties. To enhance model generalizability, data augmentation with a factor of 20 was applied, and temperature sampling at 0.8 was used to increase structural diversity. The model demonstrated a success rate of 73.1% in designing novel inhibitors. Among the generated compounds, 78 with predicted inhibitory potency below 20 µM were selected for molecular docking at the active site of GSK-3β. Docking results indicated that compound A, featuring an isoquinoline core (predicted binding energy: −5.16 kcal/mol), and compound B, featuring a quinoline core (predicted binding energy: −5.25 kcal/mol), were the most promising candidates in terms of binding affinity. Furthermore, drug-likeness assessments of both selected compounds confirmed their full compliance with pharmacokinetic rules and oral absorption criteria. This integrated approach can serve as an effective framework for the design of selective and potent inhibitors in Parkinson’s disease therapy. |
|
| Keywords: Artificial Intelligence, Chemical Language Model, GSK-3β, Parkinson’s Disease, Molecular Docking |
|
|
|
Type of Study: Research |
Subject:
Special
Received: 2025/06/8 | Accepted: 2025/09/8 | Published: 2025/09/17
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
