EnsembleDL-ATG: Identifying autophagy proteins by integrating their sequence and evolutionary information using an...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Autophagy is essential for cellular homeostasis, with autophagy-related (ATG) proteins regulating the process.
• Accurate identification of ATGs is crucial for understanding autophagy regulation.
• Existing computational methods for ATG prediction from protein sequences have limitations.

Purpose of the Study:

• To develop an advanced computational framework for predicting autophagy-related (ATG) proteins.
• To improve the accuracy and reliability of ATG identification using machine learning.
• To establish a new benchmark for ATG prediction in bioinformatics.

Main Methods:

• Proposed EnsembleDL-ATG, an ensemble deep learning framework integrating multiple models.
• Utilized protein sequence and evolutionary information as input features.
• Evaluated individual deep learning models and explored ensemble combinations.
• Developed a final framework comprising four distinct deep learning models.

Main Results:

• Achieved a prediction accuracy of 94.5% and a Matthews Correlation Coefficient (MCC) of 0.890.
• Demonstrated performance improvements of nearly 4% in accuracy and 0.08 in MCC compared to ATGPred-FL.
• Established EnsembleDL-ATG as the first ATG machine learning predictor based on ensemble deep learning.

Conclusions:

• EnsembleDL-ATG offers a significant advancement in computational prediction of autophagy-related proteins.
• The framework provides a more accurate and robust tool for autophagy research.
• The developed methods and data are publicly available to facilitate further research.