transFold

transFold predicts the super-secondary structure of transmembrane beta-barrel (TMB) proteins using pairwise inter-strand residue statistical potentials, multi-tape S-attribute grammars, and dynamic programming to identify global minimum-energy configurations.

Key Features:

• Super-secondary structure and topology prediction: Predicts the super-secondary structure and topology of transmembrane beta-barrel (TMB) proteins.
• Side-chain orientation prediction: Predicts side-chain orientations of transmembrane beta-strand residues.
• Inter-strand contact prediction: Predicts inter-strand residue contacts between beta-strands.
• Strand inclination prediction: Predicts the inclination of transmembrane beta-strands relative to the membrane.
• Statistical potentials: Uses pairwise inter-strand residue statistical potentials derived from globular (non-outer-membrane) proteins.
• Grammar-based modeling: Employs multi-tape S-attribute grammars to describe all potential TMB conformations.
• Optimization by dynamic programming: Applies dynamic programming to identify the global minimum-energy supersecondary structure.
• No conventional machine learning: Does not rely on hidden Markov models or neural networks.
• Reported accuracy: Reported as the most accurate method for predicting beta-barrel structures.

Scientific Applications:

• Structural prediction for outer membrane proteins: Predicts structures of TMB proteins found in Gram-negative bacteria, mitochondria, and chloroplasts.
• Residue-level structural analysis: Provides side-chain orientations and inter-strand contacts for residue-level studies of TMB proteins.
• Addressing limited experimental structures: Provides computational predictions to mitigate the scarcity of experimentally determined non-homologous TMB structures.
• Support for biological research: Supports research in microbiology, biochemistry, and molecular biology concerned with TMB protein structure and function.

Methodology:

Uses pairwise inter-strand residue statistical potentials derived from globular proteins, represents possible TMB conformations with multi-tape S-attribute grammars, and applies dynamic programming to select the global minimum-energy supersecondary structure; does not use hidden Markov models or neural networks.