flDPnn

flDPnn predicts intrinsic protein disorder, disordered linkers, fully disordered proteins, and disorder-associated functions from protein sequences using a deep neural network.

Key Features:

• Deep neural network: Employs an advanced deep neural network architecture for residue-level disorder prediction.
• Sequence profiles and input encoding: Leverages derived sequence profiles and specialized input encoding techniques to improve predictive accuracy.
• Comprehensive predictions: Predicts intrinsic disorder, fully disordered proteins, disordered linkers, and four common disorder functions.
• Performance validation: Demonstrated superior performance relative to existing predictors through testing including the Critical Assessment of protein Intrinsic Disorder prediction (CAID) experiment and other datasets.
• Ablation testing: Uses ablation tests that attribute performance gains to the methods for deriving sequence profiles and encoding inputs.

Scientific Applications:

• Protein function analysis: Enables inference of disorder-related cellular roles by predicting disorder functions in proteins.
• Research and development: Provides disorder and function predictions to support hypothesis generation and experimental design in protein science.
• Biomedical research: Aids studies of diseases involving misfolded or disordered proteins by characterizing disorder and functional regions.

Methodology:

Uses a deep neural network with derived sequence profiles and input encoding approaches, with performance evaluated via ablation tests and validation on datasets including the CAID experiment.