CSpritz

CSpritz predicts intrinsic protein disorder from amino acid sequences by integrating multiple prediction methods to produce consensus per-residue disorder probabilities for identifying short and long disordered regions.

Key Features:

• Punch: Utilizes sequence and structural templates trained using support vector machines to enhance disorder prediction accuracy.
• ESpritz: A single-sequence method based on bidirectional recursive neural networks for rapid disorder prediction.
• Extended Spritz: Filters predictions by considering structural homologues to improve reliability.
• Consensus integration: Averages probabilities from Punch, ESpritz, and extended Spritz to produce a robust consensus prediction.
• Prediction outputs: Provides per-residue disorder predictions for short and long disordered regions, secondary structure details, and identification of short functional linear motifs within disordered segments.

Scientific Applications:

• Disorder identification: Detection and characterization of short and long intrinsic disorder regions in proteins.
• Structural annotation: Integration of structural homologue information and secondary structure details to inform structural biology analyses.
• Benchmarking and evaluation: Performance assessed on the CASP9 dataset, reporting Sw = 49.27 and AUC = 0.828.

Methodology:

Predictions from Punch (SVMs on sequence and structural templates), ESpritz (bidirectional recursive neural networks on single sequences), and extended Spritz (filtering by structural homologues) are combined by averaging their per-residue disorder probabilities.