Assessing approaches for inferring species trees from multi-copy genes

Area of Science:

• Phylogenetics
• Computational Biology
• Genomics

Background:

• Genomic data availability fuels interest in using gene duplication and loss for species tree inference.
• Assessing nonprobabilistic and probabilistic methods for species tree reconstruction is crucial.

Purpose of the Study:

• To evaluate the performance of various species tree inference methods using gene duplication and loss simulations.
• To examine the impact of factors like gene tree error and duplication rates on phylogenetic accuracy.

Main Methods:

• Simulated gene duplication and loss (Dup-loss) and coalescence data.
• Evaluated gene tree parsimony (GTP), NJst, MulRF, and PHYLDOG.
• Assessed effects of gene sampling, gene tree error, and duplication/loss rates.

Main Results:

• MulRF outperformed other methods at low duplication rates and up to 100 taxa.
• Dup-loss was most accurate at high duplication rates and in larger simulations.
• PHYLDOG showed promise but had prohibitive run times for large datasets.
• GTP methods exhibited high error in duplication/loss estimation.

Conclusions:

• Method performance in species tree inference depends heavily on duplication and loss rates.
• Generic tree distance methods are useful for large-scale phylogenetic analyses.
• Increased gene trees and reduced missing data improve phylogenetic accuracy.