DisPredict2

DisPredict2 predicts residue-level intrinsic disorder in protein sequences using a support vector machine-based model to provide binary and probabilistic residue annotations.

Key Features:

• Support Vector Machine (SVM): Applies an SVM classifier to protein sequences provided in FASTA format to predict residue disorder.
• Position Specific Estimated Energy (PSEE): Computes PSEE from contact energy and predicted relative solvent accessibility (RSA) derived solely from sequence to identify structured and intrinsically disordered regions.
• Energy-Based Classification: Computes favorable and unfavorable energies and applies thresholds to classify residues as ordered or disordered and to assist in segregating secondary structure types.
• Residue-level Outputs: Produces binary annotations per residue (disordered/ordered) and two real-valued scores representing the probability of disorder and order.
• Hydrophobicity Correlation: PSEE values show a strong correlation with amino acid hydrophobicity.
• Detection of Disorder-to-Order Transitions: Identifies regions that undergo disorder-to-order transitions relevant to binding and signaling.

Scientific Applications:

• Intrinsic Disorder Mapping: Identifies intrinsically disordered regions implicated in enzyme regulation, signal transduction, and molecular recognition.
• Functional Site Identification: Predicts disorder-to-order transition regions to locate potential binding and regulatory sites.
• Structure–Function Analysis: Segregates secondary structure types and maps dynamic regions to inform studies of protein structure and function.

Methodology:

Accepts protein sequences in FASTA format; computes PSEE from sequence-derived contact energy and predicted RSA; calculates favorable/unfavorable residue energies and applies thresholds; and uses an SVM trained on feature sets including PSEE to produce binary residue labels and probability scores.