Identifying DNA and protein patterns with statistically significant alignments of multiple sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Sequence alignment is crucial for understanding evolutionary and functional relationships among DNA, RNA, and protein sequences.
• Scoring schemes are necessary to measure sequence relatedness, especially when sequences are not highly similar or the alignment is unknown.

Purpose of the Study:

• To develop a robust strategy for measuring sequence relatedness and determining optimal alignments for identifying functional relationships.
• To introduce a statistical framework for evaluating the significance of sequence alignments.

Main Methods:

• Utilized an information content scoring scheme based on log-likelihood.
• Developed two P-value estimation methods for information content scores, combining large-deviation statistics and numerical calculations.
• Implemented a method to count possible alignments and calculate statistical significance for comparing alignments of varying widths and sequence numbers.
• Employed a greedy algorithm for determining alignments of functionally related sequences.

Main Results:

• Presented a comprehensive approach for multiple sequence alignment, incorporating statistical significance.
• Demonstrated the accuracy of P-value calculations.
• Provided an example of using the algorithm to identify binding sites for the E. coli CRP protein.

Conclusions:

• The developed methods provide a statistically sound approach to multiple sequence alignment for functional relationship inference.
• The algorithm is effective in identifying functionally relevant sequence patterns, such as protein binding sites.