MMLigner

MMLigner aligns protein sequences using the Minimum Message Length (MML) principle to provide a statistically rigorous framework for evaluating and comparing alignment hypotheses.

Key Features:

• Statistical Framework: Uses the MML information criterion to evaluate protein sequence alignments and to determine alignment parameters systematically.
• Finite State Models with Dirichlet Distributions: Integrates finite state models and Dirichlet priors to model sequence relationships and alignment probabilities.
• Marginal Probability Landscapes: Produces marginal probability landscapes over alignment hypotheses to enable assessment of multiple competing alignments.
• Addresses Parameter Arbitrariness: Mitigates the arbitrariness of alignment parameters and the disconnect between substitution scores and gap costs.
• Performance on Distantly Related Sequences: Demonstrated effectiveness on benchmarks containing distantly related protein sequences.

Scientific Applications:

• Protein Sequence Comparison: Provides a statistically rigorous method for comparing and aligning protein sequences.
• Rationalization of Alignment Relationships: Enables researchers to rationalize and evaluate alternative alignment relationships via marginal probability landscapes.

Methodology:

Applies the Minimum Message Length information criterion together with finite state models and Dirichlet priors to compute marginal probability landscapes over possible alignment hypotheses.