Mattbench

Mattbench benchmarks protein sequence aligners using structural alignments to assess and approximate SCOP superfamily-level classification based on geometric, distance-based measures.

Key Features:

• Matt-based structural alignment: Uses the Matt structure alignment program to generate structural alignments for analysis.
• Hierarchical clustering scheme: Applies a hierarchical clustering approach to categorize protein structural domains into distinct clusters.
• Geometric distance-based metrics: Employs geometric dissimilarity and distance-based measures to quantify structural relationships.
• SCOP superfamily approximation: Approximates the Structural Classification of Proteins (SCOP) at the superfamily level using purely structural metrics.
• Benchmark set for distantly related proteins: Provides a novel, curated benchmark set specifically targeting more distantly related protein structures.
• Comparison to existing methods: Demonstrates qualitative differences in performance compared to Dali/FSSP and DaliLite.

Scientific Applications:

• Sequence aligner benchmarking: Evaluates and benchmarks the performance of protein sequence alignment algorithms using structural criteria.
• Alignment validation and refinement: Assesses and aids refinement of sequence aligners by testing against structural alignment-derived standards.
• Protein fold-space analysis: Explores organization of protein fold space and relationships among distantly related structures.
• Testing distant homology detection: Provides a framework to test aligner performance on distantly homologous protein structures.

Methodology:

Performs structural alignments with the Matt program, applies a hierarchical clustering scheme that groups domains based on geometric dissimilarity (distance-based) metrics to approximate SCOP at the superfamily level, curates a benchmark set of distantly related proteins, and compares results qualitatively to Dali/FSSP and DaliLite.