Large-Deviation Properties of Sequence Alignment of Correlated Sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Estimating the significance of DNA sequence alignment scores relies on random sequence score distributions.
• The high-scoring tail of these distributions is crucial for statistical analysis.
• Previous studies showed Gumbel distribution inaccuracies for independent and identically distributed (iid) sequences, requiring a Gaussian correction.

Purpose of the Study:

• To apply the large deviation method to local and global alignment of DNA sequences with long-range correlations.
• To investigate score distributions across the full support range.
• To determine the necessity of Gumbel distribution corrections for correlated DNA sequences.

Main Methods:

• Utilizing the large deviation method for analyzing alignment scores.
• Examining both local and global alignment types.
• Studying DNA sequences exhibiting long-range correlations.

Main Results:

• Demonstrated that a Gaussian correction to the Gumbel distribution is necessary for sequences with long-range correlations.
• Obtained probabilities as low as [Formula: see text] across the full distribution support.
• Found the Gamma distribution to be an inadequate fit for global alignments, contrary to earlier heuristic findings.

Conclusions:

• The Gumbel distribution requires a Gaussian correction for accurate modeling of DNA sequence alignment scores, especially with long-range correlations.
• The large deviation method effectively analyzes these complex distributions.
• Findings highlight the limitations of standard statistical models for real DNA sequence data.