ACP-CapsPred: an explainable computational framework for identification and functional prediction of anticancer...

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Cancer poses a significant global health threat, necessitating novel therapeutic strategies.
• Anticancer peptides (ACPs) offer a promising avenue for cancer treatment.
• In silico methods provide efficient tools for identifying and characterizing therapeutic peptides.

Purpose of the Study:

• To develop and validate a computational framework, ACP-CapsPred, for accurate identification and functional characterization of ACPs.
• To leverage advanced machine learning, specifically capsule networks, for predicting ACP activity.
• To enhance the understanding of ACPs' functional roles across diverse cancer types.

Main Methods:

• Integration of a protein language model, evolutionary information, and physicochemical properties for comprehensive peptide profiling.
• Application of capsule networks, a next-generation neural network architecture, for predictive modeling.
• A two-stage framework for distinct identification and functional characterization tasks.

Main Results:

• ACP-CapsPred achieved high accuracy (up to 95.71%) and F1-scores (up to 95.90%) in identifying ACPs on benchmark datasets.
• The framework demonstrated strong performance in characterizing ACP functional activities across five cancer types, with an average accuracy of 90.75% and F1-score of 91.38%.
• ACP-CapsPred provided interpretability, identifying key regions and residues crucial for anticancer activity.

Conclusions:

• ACP-CapsPred represents a state-of-the-art computational tool for accelerating anticancer peptide discovery and development.
• The framework offers valuable insights into the functional mechanisms of ACPs across different cancers.
• This study highlights the potential of integrating advanced computational methods for biological sequence analysis and therapeutic development.