ESpritz

ESpritz predicts intrinsically disordered regions in protein sequences to enable genome-scale and functional annotation of protein disorder.

Key Features:

• Three disorder flavors: ESpritz predicts three distinct flavors of disorder, including an NMR flexibility predictor.
• Algorithm: An ensemble of bidirectional recursive neural networks performs residue-level disorder prediction.
• Prediction modes: Provides a fast mode for genome-scale annotation without sequence profiles and a slower, profile-based mode that uses sequence profiles to increase accuracy.
• Confidence thresholds: Two selectable thresholds trade sensitivity and specificity—one maximizes disorder detection and the other limits expected false positives to 5%.
• Benchmark performance: In CASP9 ESpritz achieved an Sw of 54.82 and an AUC of 0.856, with the fast predictor approximately four orders of magnitude faster than most publicly available methods from the challenge.

Scientific Applications:

• Genomic-scale disorder annotation: Rapid annotation of entire genomes for large-scale and high-throughput studies.
• Protein function and interaction analysis: Identification of intrinsically disordered regions to inform studies of protein function and interactions mediated by flexible regions.

Methodology:

Predictions are generated by an ensemble of bidirectional recursive neural networks; the tool offers a fast non-profile mode and a slower mode that incorporates sequence profiles.