ACP-CapsPred

ACP-CapsPred employs capsule networks to identify anticancer peptides and predict their functional activities across different cancer types.

Key Features:

• Two-stage predictive framework: The first stage identifies anticancer peptides, and the second stage characterizes their functional activities across five selected cancer types.
• Multi-source feature integration: Constructs comprehensive peptide profiles by integrating a protein language model with evolutionary information and physicochemical properties.
• Capsule network architecture: Employs next-generation neural networks, specifically capsule networks, as the core predictive modeling approach.
• Model interpretability: Reveals regions and residues associated with anticancer activity, providing biological insight into peptide function.

Scientific Applications:

• Anticancer peptide identification: Distinguishes anticancer peptides from non-anticancer peptides with reported accuracies of 80.25% and 95.71% on benchmark datasets Set 1 and Set 2, respectively.
• Cancer-type-specific functional prediction: Predicts functional activities of identified anticancer peptides across five selected cancer types with an average accuracy of 90.75% and F1-score of 91.38%.
• Biological sequence analysis: Offers a novel methodological perspective applicable to other biological sequence analysis problems beyond anticancer peptides.

Methodology:

ACP-CapsPred constructs predictive models using capsule networks, integrating a protein language model with evolutionary information and physicochemical properties of peptides to build comprehensive peptide profiles.