PhoglyStruct

PhoglyStruct predicts lysine phosphoglycerylation sites from protein structural properties to support proteomic analysis of this post-translational modification and its roles in cellular function and disease.

Key Features:

• Structural information: Uses accessible surface area, backbone torsion angles, and local structure conformations to inform predictions.
• Multilayer Perceptron classifier: Employs a multilayer perceptron trained on datasets containing phosphoglycerylated and non-phosphoglycerylated lysine residues.
• Performance metrics: Reports sensitivity 0.8542, specificity 0.7597, accuracy 0.7834, Matthews correlation coefficient (MCC) 0.5468, and Area Under the Curve (AUC) 0.8077.
• Dataset files: Uses datasets stored as .mat files, specifically 'original_train' and 'original_test', for model training and testing.

Scientific Applications:

• Lysine phosphoglycerylation research: Facilitates investigation of lysine phosphoglycerylation as a post-translational modification.
• Proteome-scale prediction: Aids large-scale proteomic studies by providing predictions of phosphoglycerylation sites.
• Functional and disease studies: Supports exploration of the roles of phosphoglycerylation in cellular functions and disease mechanisms.

Methodology:

Extracts structural features (accessible surface area, backbone torsion angles, local structure conformations) and trains a multilayer perceptron classifier using .mat datasets 'original_train' and evaluates using 'original_test' containing phosphoglycerylated and non-phosphoglycerylated lysine residues.