Correspondence site: http://www.respond2articles.com/MEE/
A simple polytomy resolver for dated phylogenies
Article first published online: 21 MAR 2011
© 2011 The Authors. Methods in Ecology and Evolution © 2011 British Ecological Society
Methods in Ecology and Evolution
Volume 2, Issue 5, pages 427–436, October 2011
How to Cite
Kuhn, T. S., Mooers, A. Ø. and Thomas, G. H. (2011), A simple polytomy resolver for dated phylogenies. Methods in Ecology and Evolution, 2: 427–436. doi: 10.1111/j.2041-210X.2011.00103.x
- Issue published online: 10 OCT 2011
- Article first published online: 21 MAR 2011
- Received 13 August 2010; accepted 3 February 2011 Handling Editor: Emmanuel Paradis
Fig. S1. Inferred gamma bias for the simplest single polytomy scenarios using a Yule tree prior in BEAST rather than Birth-death tree prior. Two sizes of polytomies (100 tips and 500 tips) are shown. In both cases, there is no evidence of a bias in the pseudo-posterior distribution of gamma estimates [E(γ) = 0.0]. This is in contrast with the slight positive γ bias shown for the BD model (Fig. 1). Researchers wishing to use the Yule prior can do so by running the included ‘PolytomyResolverConstraints’ R script, and combining the output XML tags with a user generated BEAUti XML file where the tree prior is set to a Yule prior.
Figs S2–S4. Comparison between BD, EQS and RND approaches for simulated 64-tip trees. Three different sets of 10 trees were simulated (λ = 0.1, μ = 0.0, 0.05, 0.09). None of these polytomy resolution approaches appear affected by the different birth and death rates used to simulate trees. Similar to Fig. 2 from the main text, the EQS approach has a strong negative γ bias (grey). Comparison of results for these smaller trees with the 250 tip tree results shown in Figure 2, demonstrates the size-dependent bias observed in the RND approach (black). In Figs S2–S4, the RND approach better recovers the true γ values (red lines) than does the RND approach in the 250-tip trees (Figure 2). The BD approach does not demonstrate any strong bias, with the 95% confidence interval overlapping the best estimate in all the 10-tree datasets. As expected there is no bias in the imbalance estimate (Ic) for any of these three 10-tree datasets (panel B).
Fig. S5. Parameters estimated from the pseudo posterior distribution of trees resolved using the BD approach. Sixty percent of internal nodes chosen at random from the starting trees, the same 10-tree dataset presented in the main content (Fig. 4), were collapsed to polytomies. Similar to the 40% polytomized trees presented in the main content, the parameter estimates (grey) for the BD resolved tree distributions encompass the initial values (red bars) for all four parameters. No biases are apparent in γ or Ic. The mean growth rate (λ − μ) does appear to be consistently underestimated. This is likely related to the challenges of estimating a relative death rate (μ/λ).
Figs S6–S13. Parameter estimates from pseudo posterior distribution of 40% polytomized 64-tip simulated trees resolved using the BD approach. For Figs S6–S12, the BD approach was able to recover the original value for all four parameters (mean growth rate, λ − μ; relative death rate, μ/λ; Pybus’ gamma, γ; and Colless’ tree imbalance, Ic). In Figure S13, where the birth and death rates were high, the BD approach was not able to reliably estimate the birth and death rate parameters. However, there is again no observable bias in γ or Ic.
Fig. S14. Pybus’ gamma (γ) and Colless’ tree imbalance (Ic) for the mammalian supertree resolved using the BEAST birth-death (BD) method. Two pseudo-posterior tree distributions are shown, the complete set of 10 000 trees, and a subsampled set of 100 trees. Although the γ and Ic distrubtions are less smooth, there does not appear to be a difference between the full set and the subsample. The true γ and Ic values for the mammalian supertree are not known.
Fig. S15. A lineage through time plot showing both the original unresolved mammalian supertree (blue line), and the 100-tree distribution of resolved trees (grey lines, 100 tightly overlapping lines shown). It is clear from this figure that the polytomy resolution approach has a noticeable effect on the node ages, shifting nodes towards the tips as more and more polytomies are resolved. The true distribution of lineages through time is not known.
Data S1.Pseudo-posterior distribution of resolved mammalian supertrees. Mammalian supertree resolution was done using the stand-alone Polytomy Resolver script. This approach resolved all polytomies under a constant rates birth-death model. Both the complete distribution of 10,000 trees and a resampled set of 100 trees are available. Log files are available from the authors upon request.
Data S2. Polytomy Resolver scripts. See supplementary materials for detailed instructions on running the "PolytomyResolver.R" ,standalone R-script or the "PolytomyResolverConstraints.R" customizable R-script.
As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.
|MEE3_103_sm_FigS1-S15.doc||864K||Supporting info item|
|MEE3_103_sm_PolytomyResolver.R||13K||Supporting info item|
|MEE3_103_sm_PolytomyResolverConstraints.R||4K||Supporting info item|
|MEE3_103_sm_PolytomyResolverInstructions.doc||55K||Supporting info item|
|MEE3_103_sm_FritzTreers200k100treestre.zip||8136K||Supporting info item|
|MEE3_103_sm_FritzTreers2k10ktreestrees.zip||450496K||Supporting info item|
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.