Pathological Rate Matrices in Evolutionary Models
Author Information
Author(s): Harold W. Schranz, Von Bing Yap, Simon Easteal, Rob Knight, Gavin A. Huttley
Primary Institution: John Curtin School of Medical Research, The Australian National University
Hypothesis
Do pathological rate matrices exist in nature, and how do different algorithms perform in their computation?
Conclusion
The study found that the Padé with scaling and squaring algorithm is the most robust for computing non-reversible evolutionary models.
Supporting Evidence
- Pathological dinucleotide and trinucleotide matrices were found in microbial data.
- The Padé algorithm was up to 3 times faster than eigendecomposition.
- Taylor and eigendecomposition algorithms showed substantial error rates in certain matrices.
Takeaway
Some math methods used to study evolution can make big mistakes, but a new method works better and faster.
Methodology
The study compared different algorithms for computing rate matrices using protein coding gene alignments from various genomes.
Potential Biases
There was a bias towards the Padé algorithm due to its robustness in the study design.
Limitations
The study focused on specific types of sequences and may not generalize to all evolutionary models.
Participant Demographics
The study included microbial genomes and primate genomes.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website