Score distributions of gapped multiple sequence alignments down to the low-probability tail

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Mechanics

Background:

• Assessing sequence alignment significance requires score distributions of random sequences.
• Obtaining these distributions in biologically relevant high-scoring regions with exponentially small probabilities is challenging.
• Previous studies analyzed pairwise alignments, noting deviations from the Gumbel distribution for finite sequences.

Purpose of the Study:

• Extend statistical analysis to multiple sequence alignments (MSAs) with gaps.
• Investigate score distributions in biologically relevant low-probability regions (down to 10^-160).
• Compare MSA distributions with pairwise alignment distributions.

Main Methods:

• Application of statistical mechanics-based rare-event algorithms.
• Numerical computation of score distributions for global and local MSAs.
• Analysis of score distributions over a large range of probabilities.

Main Results:

• Distributions for MSAs with gaps differ from pairwise alignment distributions, even after rescaling for sequence length.
• The previously proposed Gaussian correction to the Gumbel distribution requires refinement for pairwise alignments.
• Rare-event algorithms successfully accessed extremely low probability regions.

Conclusions:

• Current statistical models for sequence alignment scores need refinement, especially for MSAs.
• Findings have significant implications for biological sequence analysis and database searching.
• Advanced computational methods are essential for accurate statistical assessments in bioinformatics.