TreeDet

TreeDet predicts subfamily-specific functional residues in protein sequence alignments by integrating tree-based phylogenetic, correlation-based distance-matrix, and PCA/clustering analyses to identify conserved positions relevant to protein function.

Key Features:

• Integration of multiple analytical methods: Combines tree-based phylogenetic representations, correlation-based distance-matrix comparisons, and PCA with clustering to detect conserved and subfamily-specific positions.
• Tree-based phylogenetic analysis: Uses phylogenetic representations to relate sequence conservation to evolutionary context.
• Correlation-based distance-matrix comparisons: Compares mutational behaviors using distance matrices to detect correlated changes among positions.
• PCA with clustering: Applies principal component analysis coupled with clustering to analyze sequence distributions in multidimensional sequence space.
• Automatic subfamily partitioning using Relative Entropy: Determines optimal subfamily divisions based on Information Theory measures, specifically Relative Entropy.
• Comparison of mutational behaviors: Compares mutational behaviors of full-length proteins with those of specific sequence positions to identify functional specificity.
• Sequence-space vector representation and dimensionality reduction: Represents proteins and residues as vectors in a generalized sequence space and projects them into lower-dimensional spaces to identify residue groups specific to subfamilies.
• Statistical validation on curated families: Validated across non-redundant protein family lists, including ligand-binding families and those with annotated functional sites, demonstrating prediction of residues near bound ligands or key functional amino acids.
• Reliability assessment of positions: Provides a reliability assessment that highlights the confidence of alignment positions used for predictions.

Scientific Applications:

• Identification of functional residues: Identifying residues critical to protein activity and specificity within protein families.
• Structural biology: Mapping predicted functional residues onto structures to interpret ligand interactions and functional sites.
• Molecular evolution studies: Investigating evolutionary patterns of conservation and divergence across subfamilies.
• Drug design: Prioritizing subfamily-specific residues that may influence ligand binding or selectivity for therapeutic targeting.
• Experimental hypothesis generation: Generating testable hypotheses about residues to validate by biochemical or mutational experiments.

Methodology:

Analyzes multiple sequence alignments using tree-based phylogenetic representations, correlation-based distance-matrix comparisons of mutational behaviors, PCA with clustering of sequence distributions, automatic subfamily partitioning via Relative Entropy, vector representation of proteins and residues in sequence space with projection to lower-dimensional spaces, comparison of full-length versus position-specific mutational behaviors, statistical validation against non-redundant protein family lists including ligand-binding and annotated functional sites, and assessment of positional reliability.