TMHcon

TMHcon predicts helix-helix contacts within transmembrane regions of membrane proteins to identify interacting alpha-helices from sequence-derived and coevolutionary features.

Key Features:

• Dual Neural Network Approach: TMHcon employs two neural networks, one trained for contacts across all transmembrane helix pairs and one specialized for contacts between non-neighboring transmembrane helices.
• Membrane-Specific Input Features: Integrates residue position within the transmembrane segment, residue orientation relative to the lipophilic environment, windowed residue profiles, and sequence distance.
• Coevolutionary Analysis: Analyzes coevolving residues in predicted transmembrane regions to identify interacting helices.
• Performance Metrics: Achieves approximately 26% accuracy in predicting contacts between residues in transmembrane segments, comparable to state-of-the-art predictors for soluble proteins.
• Membrane-Targeted Design: Represents the first contact predictor specifically developed for membrane proteins with comparable efficacy to soluble-protein contact predictors.

Scientific Applications:

• Helix Interaction Pattern Prediction: Predicts helix-helix interaction patterns to facilitate classification and prediction of membrane protein folds.
• Structural Census Compilation: Enables compilation of a structural census of membrane proteins by providing predictions of interacting alpha-helices based on coevolving residues within transmembrane regions.

Methodology:

Analyzes coevolving residues in predicted transmembrane regions and applies two neural networks that incorporate windowed residue profiles, sequence distance, residue position within the transmembrane segment, and residue orientation relative to the lipophilic environment.