Adaptive Local Realignment of Protein Sequences.

Area of Science:

• Bioinformatics
• Computational Biology
• Protein Sequence Analysis

Background:

• Multiple sequence alignment tools often use uniform parameters across entire protein sequences.
• Protein mutation rates vary significantly along residue positions, challenging standard alignment methods.

Purpose of the Study:

• To develop and evaluate a new protein sequence alignment approach that adapts to local variations in mutation rates.
• To improve the accuracy of multiple sequence alignment for proteins with diverse evolutionary histories.

Main Methods:

• Adaptive local realignment identifies low-accuracy regions and refines alignments using locally optimized parameter settings.
• The method builds upon global parameter advising by introducing adaptive local parameter adjustments.
• Candidate realignments are generated and selected based on estimated accuracy.

Main Results:

• Adaptive local realignment significantly boosts alignment accuracy, achieving up to 26% improvement over default settings on difficult benchmarks.
• The method provides a 6.4% accuracy increase compared to global advising alone.
• The approach has been implemented in the Opal aligner utilizing the Facet accuracy estimator.

Conclusions:

• Adaptive local realignment offers a substantial improvement in protein sequence alignment accuracy by accounting for local sequence characteristics.
• This adaptive strategy is particularly beneficial for hard-to-align protein families.
• The integration with existing advising methods enhances overall alignment performance.