NeuroBACE-ML: A reliability-aware screening framework for high-throughput prioritization of potent BACE1 inhibitors

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and medicinal chemistry
• Machine learning in pharmacology

Background:

• Beta-site amyloid precursor protein cleaving enzyme 1 (BACE1) is a key target in Alzheimer's disease (AD) drug discovery.
• Developing effective BACE1 inhibitors is crucial for therapeutic strategies against AD.

Purpose of the Study:

• To develop and validate NeuroBACE-ML, a reliability-aware screening framework for high-throughput prioritization of BACE1 inhibitors.
• To enhance the accuracy and reliability of virtual screening for drug candidates.

Main Methods:

• Curated human BACE1 bioactivity data from ChEMBL, standardized to a binary active/inactive scale (pIC50 ≥ 7 or ≤ 6).
• Molecules represented using Morgan fingerprints; primary classifier built using XGBoost with Optuna hyperparameter optimization.
• Incorporated probability calibration, scaffold-aware robustness, applicability domain, abstention logic, and ensemble uncertainty analysis for reliability.

Main Results:

• NeuroBACE-ML demonstrated high performance on a held-out test set (AUROC=0.986, AUPRC=0.991, MCC=0.868, balanced accuracy=0.943).
• External validation on BindingDB confirmed generalizability (AUROC=0.969, AUPRC=0.987, MCC=0.790).
• The framework successfully prioritizes high-confidence BACE1 inhibitor candidates.

Conclusions:

• NeuroBACE-ML provides a practical and deployable tool for early-stage BACE1 inhibitor candidate prioritization.
• The framework enhances reliability through multiple uncertainty assessment strategies.
• While excluding intermediate activity zones, it aids downstream medicinal chemistry and experimental validation for AD drug discovery.