MaxCluster

MaxCluster compares protein three-dimensional structures to assess structural similarity between experimental structures and low-resolution tertiary models without relying on sequence information.

Key Features:

• Sequence-independent structural alignment: MaxCluster employs a heuristic algorithm that aligns experimental protein structures and low-resolution tertiary models based on three-dimensional configuration rather than amino-acid sequence.
• Structural similarity scoring: It computes a likelihood-based structural similarity score using an extreme value distribution to model random alignments.
• Performance and benchmarking: The scoring system has been benchmarked against manual-based scores and datasets, demonstrating superior correlation with human evaluators' assessments.
• Efficiency for large-scale datasets: The algorithm is optimized for speed to support routine, large-scale all-versus-all comparisons in databases of protein models.

Scientific Applications:

• Structural genomics and protein structure prediction: Enable rapid comparison of experimental and modeled protein structures, including low-resolution models, in structural genomics projects.
• Model validation: Validate theoretical or predicted models against experimental structures by quantifying structural similarity.
• Conformational analysis: Explore and quantify conformational changes by comparing different conformations of the same protein.
• Large-scale database comparisons: Perform all-versus-all and reference-based comparisons across thousands of protein models for database-wide analyses.

Methodology:

MaxCluster applies a heuristic, sequence-independent structural alignment algorithm that aligns structures by three-dimensional configuration and computes likelihood-based similarity scores using an extreme value distribution to describe random alignments.