Comparing Coevolution Detection Methods in Proteins
Author Information
Author(s): J Gregory Caporaso, Sandra Smit, Brett C Easton, Lawrence Hunter, Gavin A Huttley, Rob Knight
Primary Institution: University of Colorado Denver
Hypothesis
Tree-ignorant methods can exhibit equivalent statistical power to tree-aware methods when controlling for shared ancestry.
Conclusion
Transformed tree-ignorant metrics can effectively control for shared ancestry and may outperform tree-aware methods in detecting coevolution.
Supporting Evidence
- Tree-ignorant methods often outperformed tree-aware methods in detecting coevolutionary signals.
- Statistical analyses confirmed that the choice of amino acid alphabet significantly affects method performance.
- Transformed metrics provided sufficient control for shared ancestry effects.
Takeaway
The study shows that some methods for finding coevolving parts of proteins can work just as well without using a family tree, and they are faster too.
Methodology
The study compared nine coevolution algorithms on real protein sequence alignments, using both tree-aware and tree-ignorant methods.
Potential Biases
The study acknowledges potential biases in the choice of alphabets and the assumptions made in evaluating coevolution.
Limitations
The performance of methods may vary based on the choice of amino acid alphabets and parameters, and the true positives for coevolution are not well defined.
Statistical Information
P-Value
1.50 × 10-3
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website