B3pred

B3pred predicts blood-brain barrier penetrating peptides (B3PPs) to enable identification and design of peptides that can cross the blood-brain barrier for therapeutic and drug-delivery applications.

Key Features:

• Prediction of B3PPs: Uses machine learning to predict peptides capable of penetrating the blood-brain barrier.
• Training dataset: Models were trained, tested, and evaluated using a dataset of BBB peptides sourced from the B3Pdb database.
• Peptide feature computation: Computes a comprehensive range of peptide features for model input.
• Feature selection: Selects relevant features to improve model accuracy, with the top 80 selected features reported.
• Machine learning models: Multiple machine learning models were developed, with the random forest algorithm identified as the most effective.
• Model performance: Random forest achieved a maximum accuracy of 85.08% and an AUROC (Area Under the Receiver Operating Characteristic curve) of 0.93 using the top 80 features.
• Protein-scanning capability: Scans protein sequences to identify regions that may function as B3PPs.
• Peptide design: Enables design of new B3PPs using analogs.

Scientific Applications:

• Therapeutic agent identification: Identifies peptides with potential to serve as therapeutic agents that can cross the BBB.
• Drug delivery enhancement: Supports selection and design of peptides to facilitate transport of drugs across the BBB.

Methodology:

Machine learning models were trained, tested, and evaluated on a BBB peptide dataset from the B3Pdb database; peptide features were computed and relevant features selected; multiple models including random forest were developed, with random forest achieving 85.08% accuracy and AUROC 0.93 using the top 80 selected features.