Revisiting BLAST's E-value
Author Information
Author(s): Lu Yang, William Stafford Noble, Uri Keich
Primary Institution: University of Waterloo, University of Washington, University of Sydney
Hypothesis
Can the E-values provided by BLAST be improved for better statistical significance in sequence alignment?
Conclusion
The study proposes a new method that provides more accurate significance analysis compared to BLAST's E-values.
Supporting Evidence
- BLAST's E-values can be overly conservative or too liberal.
- The proposed method does not rely on pre-computed parameters.
- The new approach can deliver more correct alignments in valid cases.
- SG P-values are more consistent than E-values across experiments.
Takeaway
This study looks at how well a tool called BLAST measures the similarity between DNA or protein sequences and suggests a better way to do it.
Methodology
The study evaluates BLAST's E-values and proposes an alternative significance analysis using a sample from the null distribution of random alignments.
Potential Biases
The E-values can be overly conservative or too liberal, leading to inflated significance estimates.
Limitations
The proposed method has a significant runtime penalty compared to BLAST.
Statistical Information
P-Value
0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website