EvoRator

EvoRator predicts residue-level evolutionary rates in proteins from three-dimensional (3D) protein structures using machine-learning regression models to analyze structural features.

Key Features:

• Structure-Based Predictions: Uses three-dimensional (3D) protein structures as primary input to estimate site-specific evolutionary rates independent of extensive homologous sequence data.
• Machine Learning Integration: Employs machine-learning regression algorithms to learn associations between 3D structural features and residue-level evolutionary rates.
• Orphan Proteins: Provides evolutionary-rate predictions for orphan proteins lacking homologous sequences by relying on structural information.
• Comparison of Evolutionary Rates: Facilitates direct comparison between structure-based and phylogeny-based estimates to distinguish conservation driven by function versus structure.
• Handling Gapped Sequence Alignments: Addresses gapped sequence alignments and clade-specific insertions by utilizing known 3D structures at those positions for rate estimation.
• Physicochemical Baseline Comparison: Demonstrates improved performance relative to traditional predictors based on physicochemical properties such as relative solvent accessibility and weighted contact number.

Scientific Applications:

• Orphan Proteins: Predicts evolutionary rates for proteins with few or no homologs to inform structural and functional analyses.
• Functional vs. Structural Conservation: Enables identification of sites conserved due to functional constraints rather than structural constraints via comparison of different rate estimates.
• Complex Alignments: Applicable to analyses of proteins with clade-specific insertions or deletions by accommodating gapped alignments using structural data.

Methodology:

Uses three-dimensional protein structures as input and applies machine-learning regression algorithms to learn associations between structural features and site-specific evolutionary rates; supports comparison to phylogeny-based estimates and uses known 3D structures to handle gapped alignments.