BrainPepPass: A Framework Based on Supervised Dimensionality Reduction for Predicting Blood-Brain Barrier-Penetrating...

Area of Science:

• Biotechnology
• Computational Biology
• Neuroscience

Background:

• Peptides crossing the blood-brain barrier (BBB) are crucial for understanding brain pathologies and developing treatments.
• Computational methods offer a cost-effective way to screen large peptide libraries for BBB penetration.
• Identifying BBB-penetrating peptides (B3PPs) is key for targeted drug delivery to the brain.

Purpose of the Study:

• To develop and validate a machine learning framework, BrainPepPass, for predicting B3PPs.
• To assess the performance of BrainPepPass against existing classification methods.
• To provide a publicly accessible tool for academic research on B3PPs.

Main Methods:

• Utilized supervised manifold dimensionality reduction and extreme gradient boosting (XGB) algorithms.
• Developed a machine learning (ML) framework named BrainPepPass.
• Evaluated the model using 10-fold cross-validation, leave-one-out cross-validation (LOOCV), and independent tests.

Main Results:

• BrainPepPass achieved high prediction accuracies, exceeding 94% in 10-fold CV and 98% in LOOCV.
• Independent tests demonstrated accuracy values ranging from 45% to 97.05%.
• The proposed ML framework outperformed other existing classifiers in predicting B3PPs.

Conclusions:

• BrainPepPass is a highly accurate and effective tool for predicting both natural and chemically modified B3PPs.
• The computational approach significantly advances the identification of peptides for therapeutic applications targeting the brain.
• The BrainPepPass tool is available on GitHub, facilitating further research in neurotherapeutics and drug delivery.