ATGPred-FL: sequence-based prediction of autophagy proteins with feature representation learning

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Autophagy is a critical biological process regulated by numerous autophagy proteins.
• Accurate identification of autophagy proteins is essential for understanding their functions.
• Experimental methods for protein identification are costly and labor-intensive.

Purpose of the Study:

• To develop an automated, accurate, and reliable sequence-based computational tool for identifying autophagy proteins.
• To enable efficient identification of novel autophagy proteins from large datasets.
• To provide a valuable resource for autophagy research.

Main Methods:

• Investigated various sequence-based feature descriptors for protein identification.
• Employed feature learning to generate informative probability features.
• Utilized a two-step feature selection strategy to optimize feature sets.
• Developed a support vector machine classifier for the final predictor (ATGPred-FL).

Main Results:

• Achieved high accuracy in identifying autophagy proteins: 94.40% on the training set and 90.50% on the testing set.
• Identified a discriminative 14-dimensional feature set for prediction.
• ATGPred-FL is the first machine learning predictor for autophagy proteins based on primary sequences.

Conclusions:

• ATGPred-FL is an effective and useful tool for autophagy protein identification.
• The predictor facilitates the discovery of novel autophagy proteins, aiding biological research.
• The tool, source code, and datasets are publicly available for the scientific community.