A BLAST from the past: revisiting blastp's E-value

Area of Science:

• Bioinformatics
• Computational Biology
• Genomic Research

Background:

• The Basic Local Alignment Search Tool (BLAST) is crucial for genomic research and sequence similarity searches.
• BLAST's E-value provides statistical significance for alignments, indicating expected random alignments of similar scores.

Purpose of the Study:

• To critically evaluate the accuracy of E-values from protein BLAST (blastp).
• To propose and validate an alternative statistical method for assessing alignment significance.

Main Methods:

• Evaluation of blastp E-values for potential conservativeness or liberality.
• Development of a novel significance analysis based on sampling the null distribution of random alignments.
• Comparison of the proposed method's performance against blastp E-values across extensive experiments.

Main Results:

• Blastp E-values can be inconsistently conservative or liberal.
• The proposed method provides valid significance estimates without inflated scores.
• The new approach is often less conservative than blastp E-values and can yield more correct alignments.
• The method supports flexible choices of substitution matrices and gap penalties.

Conclusions:

• The developed method offers a more reliable alternative to blastp E-values for assessing sequence alignment significance.
• This approach enhances the accuracy of identifying biologically relevant alignments in genomic research.
• The flexibility in parameters and potential for error rate control offer advantages over standard BLAST analysis.