Integrated Machine Learning and Molecular Dynamics Framework Identifies Novel GSK-3β Inhibitor Candidates for Alzheimer's Disease
核心洞察
A two-stage machine learning virtual screening framework achieved significantly higher balanced accuracy (0.86 vs. 0.74) over standard multiclass models for identifying GSK-3β inhibitors.
No FDA-approved GSK-3β inhibitors currently exist for Alzheimer's disease (搜索) due to challenges in isoform selectivity, safety, and pharmacokinetic limitations.
Top hit compounds showed binding affinity profiles consistent with potential CMGC family Multi-Target Directed Ligands, advantageous for mitigating multiple kinase pathways driving Tau hyperphosphorylation.
A research team has developed an integrated computational framework combining machine learning, molecular docking, and molecular dynamics simulations to identify novel Glycogen Synthase Kinase-3 Beta (搜索) (GSK-3β) inhibitors for Alzheimer's disease (搜索) (AD) therapy. The study, published in Scientific Reports, addresses a critical gap in the AD treatment landscape: no FDA-approved GSK-3β inhibitors are currently available despite the enzyme's well-established role in driving Tau hyperphosphorylation, a hallmark of Alzheimer's pathology.
The absence of approved GSK-3β-targeting therapies stems from persistent challenges in isoform selectivity, safety profiles, and pharmacokinetic limitations. This new research tackles these obstacles through a multi-pronged computational strategy.
Two-Stage Machine Learning Outperforms Standard Approaches
The investigators developed an OECD guideline-compliant, two-stage machine learning-based virtual screening framework. A chemically diverse dataset sourced from multiple databases was pre-processed and used for model development and validation. A comparative analysis demonstrated the superiority of this two-stage approach over standard multiclass models, yielding significantly higher balanced accuracy on the internal test set — 0.86 compared to 0.74 — along with improved specificity.
The best predictive models were subsequently deployed as an open-access web tool, enabling broader scientific community access. These models were then applied to screen the ASINEX Synergy Library (搜索) for potential GSK-3β inhibitors.
Structure-Based Validation and Multi-Target Potential
In the structure-based phase, molecular docking was performed using a validated docking protocol. The top-performing molecules from this screen advanced to molecular dynamics simulations, binding free energy calculations, and per-residue decomposition analysis. Principal component analysis of the simulation trajectories confirmed global stability and consistent binding modes across the candidate compounds.
To evaluate selectivity, the researchers conducted cross-screening against the homologous GSK-3α isoform and the structurally distinct Cyclin-dependent kinase 2 (搜索) (CDK2) through molecular docking. Rather than demonstrating strict single-target exclusivity, the top hits exhibited binding affinity profiles consistent with potential CMGC family Multi-Target Directed Ligands (MTDLs).
"This presumed polypharmacological profile is advantageous for Alzheimer's therapeutics, positioning these compounds as robust candidates for simultaneously mitigating multiple kinase pathways that drive Tau hyperphosphorylation," the authors note.
A Reproducible Path Forward
The integrated workflow — spanning machine learning model development, validation, screening, protein selection, molecular docking, and molecular dynamics — provides what the researchers describe as "a reproducible, interpretable, and high-confidence method for identification of GSK-3β inhibitors for Alzheimer's disease (搜索)." The authors emphasize that experimental kinome validation is warranted to confirm the predicted polypharmacological profiles of the lead candidates.
The study represents a significant computational advance in the search for disease-modifying Alzheimer's therapies, offering both methodological innovation and tangible chemical starting points for future drug development efforts targeting GSK-3β-driven neurodegeneration.