AAGP integrates physicochemical and compositional features for machine learning-based prediction of anti-aging...

Area of Science:

• Biotechnology
• Computational Biology
• Gerontology

Background:

• Aging is a natural process involving physiological decline, leading to conditions like wrinkles and immune dysfunction.
• Peptide therapies offer a promising avenue for anti-aging interventions due to their safety and specificity.
• Computational approaches can accelerate the identification of effective anti-aging peptides.

Purpose of the Study:

• To develop and validate an accurate computational tool for predicting anti-aging peptides.
• To explore the utility of physicochemical and compositional features in anti-aging peptide prediction.
• To provide an open-source tool (AAGP) for researchers in the field.

Main Methods:

• Construction of two distinct datasets with anti-aging peptides as positive examples.
• Encoding peptides using 4,305 features and employing adaptive feature selection.
• Utilizing nine machine learning models for rigorous cross-validation and independent testing.

Main Results:

• The AAGP predictor demonstrated strong performance, achieving MCCs of 0.692 and 0.580, and AUCs of 0.963 and 0.808 on independent test sets.
• Feature importance analysis revealed that physicochemical features were more critical for one dataset, while compositional features dominated the other.
• The study successfully identified key features driving anti-aging peptide prediction.

Conclusions:

• AAGP provides a reliable computational method for identifying potential anti-aging peptides.
• The findings highlight the differential importance of peptide features in predicting anti-aging activity.
• The availability of AAGP's source code facilitates further research and development in anti-aging peptide discovery.