Prediction and analysis of anti-aging peptides using data augmentation and machine learning algorithms.

Area of Science:

• Biogerontology
• Computational Biology
• Peptide Therapeutics

Background:

• Aging is a natural process leading to age-related diseases, posing healthcare challenges.
• Anti-aging peptides (AAPs) show promise for therapeutic interventions due to their favorable properties.
• A lack of computational tools impedes systematic research and discovery of AAPs.

Purpose of the Study:

• To develop and validate novel computational models for predicting anti-aging peptides (AAPs).
• To establish a benchmark dataset for AAP research.
• To accelerate the identification and development of AAPs for therapeutic applications.

Main Methods:

• Created a benchmark AAP dataset using annotated biological functions and the AgingBase database.
• Developed three predictive models: Antiaging-FL (feature learning + ML), ESM_GAN (GAN for data augmentation), and ESM_CNN (CNN with data augmentation).
• Employed machine learning algorithms, generative adversarial networks, and convolutional neural networks for peptide prediction.

Main Results:

• All three proposed models demonstrated high predictive performance.
• Antiaging-FL achieved an AUC of 1.00 on the AAP400 dataset and 0.99 on an independent test set.
• ESM_GAN and ESM_CNN also showed strong performance with AUCs of 0.99/0.95 and 0.96/0.94, respectively.

Conclusions:

• The study presents three effective computational models for AAP prediction, aiding in the discovery of novel anti-aging peptides.
• These models serve as valuable resources for researchers and offer insights into AAP mechanisms.
• The findings facilitate targeted drug development for aging-related conditions.