ATGPred-FL

ATGPred-FL predicts autophagy proteins from protein primary sequences to enable accurate identification of autophagy-related proteins for downstream biological and disease-focused studies.

Key Features:

• Sequence-based feature representation learning: Utilizes various sequence-based feature descriptors and a feature learning method to generate probability features that enhance prediction accuracy.
• Two-step accuracy-based feature selection: Applies an accuracy-based two-step feature selection strategy to remove irrelevant and redundant features, yielding a discriminative 14-dimensional feature set.
• Support vector machine classifier: Trains an SVM classifier that achieved 94.40% accuracy on the training set and 90.50% accuracy on the testing set.

Scientific Applications:

• Autophagy protein identification from sequences: Enables identification of autophagy proteins from protein and peptide primary-sequence datasets.
• Investigation of biological roles and disease mechanisms: Facilitates exploration of autophagy protein functions and studies of diseases associated with dysfunctional autophagy.

Methodology:

Investigation of sequence-based feature descriptors; application of a feature learning method to derive probability features; implementation of an accuracy-based two-step feature selection strategy producing a 14-dimensional feature set; and support vector machine classification.