Protein domain hierarchy Gibbs sampling strategies

Area of Science:

• Computational Biology
• Bioinformatics
• Evolutionary Biology

Background:

• Hierarchically arranged multiple sequence alignment profiles aid in modeling protein domain functional divergence.
• Current methods for constructing these hierarchies, like those for the NCBI Conserved Domain Database (CDD), rely heavily on manual curation.

Purpose of the Study:

• To introduce an automated method for constructing hierarchical alignment profiles using a Gibbs sampler.
• To identify sequence determinants of protein function through statistical evolutionary dynamics analysis.
• To characterize ambiguities and derive robust consensus models for protein domain evolution.

Main Methods:

• Development and application of a Gibbs sampler employing statistical evolutionary dynamics analysis.
• Testing the sampler on simulated protein sequences conforming to the statistical model.
• Application to real protein sequences to assess hierarchy convergence and identify ambiguities.

Main Results:

• The Gibbs sampler efficiently converges on the correct hierarchy for simulated data.
• For real protein sequences, the sampler identifies multiple, nearly-optimal hierarchies, indicating significant ambiguity.
• Methods are presented to analyze these ambiguities and determine robust consensus hierarchy features.

Conclusions:

• Automated construction of hierarchical alignment profiles is feasible using statistical evolutionary dynamics.
• Real protein domain evolution presents inherent ambiguities in hierarchical modeling.
• Consensus hierarchies derived from ensembles provide stable models for understanding protein domain functional divergence.