Matt

Matt performs multiple protein structure alignment by incorporating local flexibility between structural fragments to improve alignment accuracy and detect distant homologs.

Key Features:

• Aligned fragment pair chaining algorithm: Chains aligned fragment pairs and introduces small translations and rotations during intermediate steps to enhance proximity of aligned fragments even when temporarily violating rigid body transformations.
• Local flexibility modeling: Permits "bent" alignments and temporary backbone distortions to capture local flexibility between structural fragments.
• Short-fragment optimal alignment: Begins by optimally aligning short structural fragments from multiple proteins.
• Dynamic programming assembly: Uses dynamic programming to assemble fragments into a coherent alignment guided by bent alignments.
• Geometric consistency restoration: Restores geometric consistency in a final output step after using permissive intermediate adjustments.
• p-value scoring: Computes a p-value score derived from common-core length and average root mean squared deviation (RMSD) to assess significance of alignments.
• Helix and strand end alignment: Enhances alignment of alpha-helix and beta-strand termini, improving template construction for threading approaches.
• Performance on benchmarks: Demonstrates competitive global performance on Homstrad and superior performance on SABmark, particularly for distant homologs.
• Modeling conformational states: Accommodates proteins in various conformational states to represent structural variation across homologs.

Scientific Applications:

• Multiple protein structure alignment: Produces alignments that account for local flexibility to improve structural comparison across protein sets.
• Distant homology detection: Distinguishes distantly homologous structure pairs from non-homologous proteins using p-value scores based on common-core length and average RMSD.
• Structural template library construction: Aligns helix and strand ends to aid construction of structural template libraries for threading and the inverse protein-folding problem.
• Conformational variability analysis: Analyzes backbone distortions and different conformational states among related proteins.
• Benchmark evaluation: Enables performance assessment using benchmark datasets such as Homstrad and SABmark.

Methodology:

Matt initially optimally aligns short structural fragments, chains them via an aligned fragment pair chaining algorithm using small translations and rotations and dynamic programming assembly informed by "bent" alignments, then restores geometric consistency in the final output; significance is assessed via a p-value from common-core length and average RMSD.