The probabilities of similarities in DNA sequence comparisons

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Identifying significant local similarities in DNA sequences is fundamental for understanding gene function and evolutionary relationships.
• Existing methods, including some commercial packages, may not accurately assess the statistical significance of these similarities.
• The Queen and Korn algorithm provides a framework for detecting local sequence alignments.

Purpose of the Study:

• To statistically evaluate the significance of local DNA sequence similarities.
• To illustrate a procedure for assessing significance using the Queen and Korn algorithm.
• To provide a practical tool (a table) for determining the significance of similarity lengths.

Main Methods:

• Defined statistical significance at the 5% level based on the probability of observing a similarity length or greater in random sequences.
• Related the distribution of longest similarity lengths to lengths from specific starting positions.
• Implemented the Queen and Korn algorithm, constructing distributions by combining base blocks for similarity extension.

Main Results:

• Developed a method to calculate critical values for assessing the statistical significance of DNA sequence similarities.
• Generated a table to evaluate the significance of longest similarities in sequences up to 1000 bases.
• Demonstrated that substantial similarity lengths can occur by chance alone.

Conclusions:

• The calculated critical values offer a more reliable assessment of significance compared to expected numbers used in some software.
• The proposed method and accompanying table enhance the accuracy of identifying biologically meaningful DNA similarities.
• This approach aids researchers in distinguishing true biological signals from random occurrences in DNA sequence comparisons.