MoRFMPM

MoRFMPM predicts molecular recognition features (MoRFs) in protein sequences to identify intrinsically disordered segments that undergo disorder-to-order transitions and mediate protein–protein interactions.

Key Features:

• Sequence-Based Prediction: Employs a sequence-based approach that does not rely on external predictors or separate calculations of surrounding region properties.
• Minimax Probability Machine (MPM): Uses a minimax probability machine for classification, which is suited to imbalanced datasets.
• Feature Utilization: Utilizes 16 distinct features and evaluates three different window sizes to inform predictions.
• Linear Classification Algorithm: Implements a linear classification algorithm as the classifier architecture.
• Preprocessing: Applies preprocessing to refine feature inputs prior to classification.
• Comparative Performance: Achieves higher Area Under the Curve (AUC) and maintains high True Positive Rate (TPR) at low False Positive Rate (FPR) compared with ANCHOR, MoRFpred, and MoRF_CHiBi.

Scientific Applications:

• Protein–Protein Interaction Mapping: Identifies MoRFs that mediate protein–protein interactions to support mapping of interaction interfaces.
• Study of Intrinsically Disordered Proteins: Facilitates analysis of intrinsically disordered regions and their functional roles.
• Disease Mechanism Investigation: Supports investigation of disease mechanisms involving disordered proteins.
• Drug Discovery: Assists drug discovery efforts targeting protein–protein interaction sites by locating potential binding segments.

Methodology:

Processes protein sequences using 16 predefined features across three window sizes, applies preprocessing to refine feature inputs, and classifies residues using a minimax probability machine implemented as a linear classifier, with performance evaluated by AUC and TPR at low FPR against ANCHOR, MoRFpred, and MoRF_CHiBi.