TAXFOLD

TAXFOLD performs protein fold recognition by classifying query amino acid sequences into known structural folds using taxonomy-based evolutionary context and profile-derived structural features.

Key Features:

• Taxonomy-based classification: Classifies proteins based on evolutionary relationships to incorporate phylogenetic context into fold recognition.
• Sequence-evolution profiles (PSI-BLAST): Uses PSI-BLAST-derived profiles to capture evolutionary patterns and conserved motifs from amino acid sequences.
• Secondary-structure profiles (PSIPRED): Integrates PSIPRED secondary-structure predictions to represent local structural elements relevant to folding.
• Comprehensive 137-feature descriptor set: Constructs 137 features representing global and local characteristics of PSI-BLAST and PSIPRED profiles for discriminative representation.
• Extraction of discriminative sequence features: Focuses on extracting highly discriminative features from amino acid sequences for accurate fold discrimination.

Scientific Applications:

• Benchmark fold recognition: Achieves recognition accuracies of 79.6%–90% across fold sets of 27, 95, and 194, with an average improvement of 6.9% over the best taxonomic methods.
• Template-based threading comparison: Produces results on the Lindahl benchmark dataset that are comparable to conventional template-based threading methods at the SCOP fold level.

Methodology:

Taxfold derives PSI-BLAST profiles and PSIPRED secondary-structure profiles, constructs a 137-feature representation from those profiles, and applies a taxonomy-based classification using evolutionary relationships.