NorsNet

NorsNet predicts natively unstructured (disordered) regions within protein sequences to identify long contiguous segments lacking regular secondary structure (NORS) for applications in protein annotation and structural genomics.

Key Features:

• Neural network-based approach: Employs a neural network methodology that targets nonregular secondary structure segments termed NORS (Non-Ordered Regular Structure).
• Training on predicted information: Trained using predicted rather than experimental data to avoid training/testing overlap and to enable comprehensive proteome coverage.
• Loop distinction: Focuses on distinguishing unstructured loops from well-structured loops by exploiting substantial structural differences between them.
• Eukaryotic prevalence: Explicitly addresses regions that are particularly prevalent in eukaryotic proteins.
• Benchmark validation: Benchmarked against previously used and novel experimental data to confirm predictive effectiveness.
• Comparative analysis: Compared with methods such as DISOPRED2 and shown to identify long unstructured loops as a major component of disordered regions and to detect unstructured regions at domain boundaries more often than expected by chance.

Scientific Applications:

• Protein annotation: Flags previously unannotated unstructured regions within proteins to improve functional and structural annotation.
• Structural genomics: Identifies unstructured regions in putative structural genomics targets to inform target selection and expand structural coverage of large eukaryotic families.
• Protein–protein interaction analysis: Supports estimates that approximately 50%–70% of proteins with more than seven interaction partners contain unstructured regions, implicating these regions in molecular networks.

Methodology:

Uses a neural network trained on predicted secondary structure information to distinguish NORS from structured loops, validated by benchmarks on experimental data and compared to DISOPRED2.