LightBBB

LightBBB predicts blood-brain barrier (BBB) permeability of chemical compounds to support selection of neurotherapeutic candidates.

Key Features:

• Algorithm: Uses the Light Gradient Boosting Machine (LightGBM) algorithm for model training and prediction.
• Dataset: Trained on 7,162 compounds with known BBB permeability, comprising 5,453 BBB-permeable (BBB+) and 1,709 non-permeable (BBB-) compounds.
• Performance Metrics: Achieves 89% overall accuracy, area under the curve (AUC) of 0.93, specificity of 77%, and sensitivity of 93% via 10-fold cross-validation.
• Validation: Externally validated on 74 central nervous system (CNS) compounds from the literature, yielding 90% accuracy, 85% sensitivity, and 94% specificity.

Scientific Applications:

• Early-stage screening: Rapidly screens chemical libraries to prioritize compounds for BBB permeability assessment during early brain drug discovery.
• Candidate prioritization: Supports selection of promising neurotherapeutic candidates by predicting likelihood of BBB penetration, reducing reliance on experimental assays.

Methodology:

Compiled a literature-derived dataset of 7,162 compounds and trained a LightGBM model, evaluated by 10-fold cross-validation and externally validated on 74 CNS compounds.