Stack-AAgP: Computational prediction and interpretation of anti-angiogenic peptides using a meta-learning framework

Area of Science:

• Biochemistry
• Computational Biology
• Oncology

Background:

• Angiogenesis is vital in human diseases, especially solid tumors.
• Peptides with anti-angiogenic properties offer a promising cancer treatment avenue.
• Accurate identification of anti-angiogenic peptides (AAPs) is key for drug discovery.

Purpose of the Study:

• To develop a novel ensemble-learning model for accurate identification and interpretation of AAPs.
• To leverage machine learning algorithms and feature encoding for enhanced AAP prediction.

Main Methods:

• A novel ensemble-learning model, Stack-AAgP, was developed.
• 24 baseline models were generated using six ML algorithms and four feature encodings.
• A stacked ensemble framework integrated meta-classifiers for the final predictive model.

Main Results:

• Stack-AAgP significantly outperforms state-of-the-art methods.
• Evaluated on the NT15 dataset, Stack-AAgP showed accuracy improvements of 5%-7.5%.
• Matthews Correlation Coefficient (MCC) increased by 7.2%-12.2% compared to existing predictors.

Conclusions:

• Stack-AAgP demonstrates superior performance in identifying AAPs.
• The model provides a robust framework for discovering novel anti-cancer drugs.
• Further research can explore hyperparameter optimization for Stack-AAgP.