PPLK+C

PPLK+C predicts peptide ligands of potassium channels from primary sequence to identify channel-specific peptides for basic research and therapeutic development.

Key Features:

• Primary-sequence-based prediction: Predicts peptide ligands using only primary amino acid sequence information.
• Amino acid molecular features: Leverages multiple amino acid molecular features for model input.
• Implementation: Implemented in Python.
• Machine-learning algorithms: Employs random forest, nearest neighbors, support vector machines, and artificial neural networks.
• Training and validation: Models are trained and validated using biological data derived from peptides with experimentally verified activity.
• Performance (random forest): Reported sensitivity 0.77, specificity 0.94, accuracy 0.91, and Matthews correlation coefficient 0.70 for the random forest model.

Scientific Applications:

• Peptide ligand discovery: Identification of candidate peptide ligands for potassium channels from sequence data.
• Therapeutic development: Prioritization of peptide candidates relevant to diseases such as cardiovascular disorders and cancer.
• Channel–peptide interaction studies: Support for investigating interactions between peptides and potassium channels.

Methodology:

Implemented in Python; extracts amino acid molecular features and trains four machine-learning algorithms—random forest, nearest neighbors, support vector machines, and artificial neural networks—using datasets of peptides with experimentally verified activity, with reported performance metrics for the random forest model (sensitivity 0.77, specificity 0.94, accuracy 0.91, MCC 0.70).