TopProperty

TopProperty predicts per-residue structural and topological properties of proteins, providing secondary structure, solvent accessibility, transmembrane topology, and membrane exposure annotations for transmembrane proteins (TMPs) and globular proteins.

Key Features:

• Per-residue predictions: Provides residue-level annotations for secondary structure, solvent accessibility, transmembrane topology, and membrane exposure.
• Integrator of multiple predictors: Leverages the combined outputs of 27 primary predictors.
• Deep learning ensembles: Uses two ensembles of deep neural networks to integrate predictor outputs and generate consensus predictions.
• Protein scope: Applicable to both transmembrane proteins (TMPs) and globular proteins.
• Training data bias control: Trained on datasets curated to avoid bias towards sequences with a high number of homologs.
• Unified prediction capability: Provides comprehensive per-residue annotations that remove the need for separate specialized TMP prediction tools.

Scientific Applications:

• Experimental design: Informs design and interpretation of experiments by supplying residue-level topology and exposure information for TMPs and soluble proteins.
• Protein structure prediction: Supplies per-residue constraints and annotations useful for model building and validation in structure prediction workflows.
• Transmembrane protein analysis: Supports analysis of TMPs that are underrepresented in experimentally resolved structures by predicting topology and membrane exposure.

Methodology:

Integrates outputs from 27 primary predictors via two ensembles of deep neural networks and is trained on datasets curated to avoid bias toward sequences with many homologs to produce per-residue predictions of secondary structure, solvent accessibility, transmembrane topology, and membrane exposure.