MATRAS

MATRAS compares three-dimensional protein structures using a Markov transition model to quantify structural similarity and infer evolutionary relationships.

Key Features:

• Markovian Structure Similarity Scoring: Computes a similarity score defined as log P(j → i)/P(i) based on a Markov transition model that evaluates the probability of one structure evolving into another to detect homologous structural similarities.
• Hierarchical Alignment Algorithm: Performs hierarchical alignment beginning with a rough alignment of secondary structure elements (SSEs) followed by refinement using more detailed scoring functions.
• Multiple Similarity Scores: Produces Environment Score, Residue-Residue Distance Score, and Secondary Structure Elements (SSE) Score to assess complementary aspects of structural similarity.
• Progressive Multiple-Structure Alignment: Implements a progressive alignment algorithm that assembles pairwise alignments in an optimal sequence for multiple 3D alignment.

Scientific Applications:

• Homology Detection: Identification of homologous protein structures through probabilistic structural similarity scoring.
• Evolutionary Inference: Inference of evolutionary relationships among proteins based on estimated structural transition probabilities.
• Functional Annotation Support: Support for functional inference by recognizing structurally similar proteins that may share function.
• Superfamily Recognition and Benchmarking: Recognition of superfamily relationships validated by comparison with the SCOP database and assessed relative to methods such as FSSP.

Methodology:

Estimate transition probabilities using a Markov model (akin to Dayhoff's amino acid substitution model); identify homologous protein-structure pairs via sequence similarity; analyze frequencies of structural transitions within these pairs to construct a transition probability matrix; derive longer-term transition probabilities by iterative multiplication; define similarity as log P(j → i)/P(i).