DISOPRED3

DISOPRED3 predicts intrinsically disordered regions (IDRs) in eukaryotic proteins and annotates potential protein-binding sites within those disordered regions to support studies of protein disorder and function.

Key Features:

• IDR prediction: Predicts intrinsically disordered regions in eukaryotic protein sequences.
• Protein-binding site annotation: Identifies potential protein-binding sites within predicted disordered regions.
• Machine-learning methods: Employs advanced machine-learning methodologies to improve prediction accuracy.
• SVM-based classifier: Uses a Support Vector Machine classifier to assign protein-binding annotations to disordered regions.
• Profile and sequence-derived features: Leverages sequence profile data alongside additional sequence-derived features as input to the classifier.
• Improved performance for long regions: Demonstrates increased specificity and sensitivity relative to DISOPRED2, particularly for regions longer than 20 amino acids.
• Benchmark validation: Performance has been validated through CASP evaluation results and full cross-validation benchmarking.

Scientific Applications:

• Functional annotation of proteins: Linking IDRs and predicted disordered binding sites to biological function in eukaryotic proteins.
• Protein structure prediction benchmarking: Contributing validation data and comparative performance metrics in CASP and related assessments.
• Study of protein disorder at multiple scales: Supporting analyses of protein disorder and function at molecular and systems biology levels.

Methodology:

Employs machine-learning approaches including a Support Vector Machine classifier that uses sequence profile data and additional sequence-derived features, with performance assessed by full cross-validation and CASP benchmarking.