Related Experiment Videos
Improved bootstrap confidence limits in large-scale phylogenies, with an example from Neo-Astragalus (Leguminosae)
M J Sanderson1, M F Wojciechowski
1Section of Evolution and Ecology, University of California, Davis, California 95616, USA. mjsanderson@ucdavis.edu
Systematic Biology
|July 16, 2002
Summary
Phylogenetic analyses show that bootstrap support for clades can decrease with more taxa. This study suggests statistical bias, not just computational limits, explains this trend in large datasets.
Area of Science:
- Phylogenetics
- Computational Biology
- Plant Systematics
Background:
- Phylogenetic analyses of large datasets present challenges.
- Bootstrap support for clades may decline with increased taxon sampling.
Purpose of the Study:
- To document the decline in bootstrap support with increasing taxon sampling in the angiosperm genus Astragalus.
- To investigate explanations for this observed decline in phylogenetic analyses.
Main Methods:
- Phylogenetic analyses using parsimony and neighbor-joining methods.
- Examination of heuristic search algorithms and bootstrap support.
- Evaluation of statistical bias and computational limitations as explanations.
Main Results:
- Bootstrap support for the Neo-Astragalus subclade declined monotonically with increased taxon sampling.
- This decline was observed irrespective of the phylogenetic method or heuristic search algorithm used.
- Statistical bias from random homoplasy appears to be the primary explanation, with computational shortcomings also contributing.
Conclusions:
- Bootstrap proportions in current phylogenetic analyses may not accurately reflect confidence assessments for large datasets.
- Iterated bootstrap methods, assessing confidence as type I error levels, can mitigate the decline in confidence with increasing taxon numbers.