Implementing and testing the multispecies coalescent model: A valuable paradigm for phylogenomics


Edwards SV, Xi Z, Janke A, Faircloth BC, McCormack JE, Glenn TC, Zhong B, Wu S, Lemmon EM, Lemmon AR. Implementing and testing the multispecies coalescent model: A valuable paradigm for phylogenomics. Molecular Phylogenetics and Evolution. 2016;94 :447-462.


In recent articles published in
Molecular Phylogenetics and Evolution
, Mark Springer and John Gatesy (S&G)
present numerous criticisms of recent implementations and testing of the multispecies coalescent (MSC)
model in phylogenomics, popularly known as ‘‘species tree” methods. After pointing out errors in align-
ments and gene tree rooting in recent phylogenomic data sets, particularly in Song et al. (2012) on mam-
mals and Xi et al. (2014) on plants, they suggest that these errors seriously compromise the conclusions
of these studies. Additionally, S&G enumerate numerous perceived violated assumptions and deficiencies
in the application of the MSC model in phylogenomics, such as its assumption of neutrality and in par-
ticular the use of transcriptomes, which are deemed inappropriate for the MSC because the constituent
exons often subtend large regions of chromosomes within which recombination is substantial. We
acknowledge these previously reported errors in recent phylogenomic data sets, but disapprove of
S&G’s excessively combative and taunting tone. We show that these errors, as well as two nucleotide
sorting methods used in the analysis of
, have little impact on the conclusions of those papers.
Moreover, several concepts introduced by S&G and an appeal to ‘‘first principles” of phylogenetics in an
attempt to discredit MSC models are invalid and reveal numerous misunderstandings of the MSC.
Contrary to the claims of S&G we show that recent computer simulations used to test the robustness
of MSC models are not circular and do not unfairly favor MSC models over concatenation. In fact, although
both concatenation and MSC models clearly perform well in regions of tree space with long branches and
little incomplete lineage sorting (ILS), simulations reveal the erratic behavior of concatenation when sub-
jected to data subsampling and its tendency to produce spuriously confident yet conflicting results in
regions of parameter space where MSC models still perform well. S&G’s claims that MSC models explain
little or none (0–15%) of the observed gene tree heterogeneity observed in a mammal data set and that
MSC models assume ILS as the only source of gene tree variation are flawed. Overall many of their crit-
icisms of MSC models are invalidated when concatenation is appropriately viewed as a special case of the
MSC, which in turn is a special case of emerging network models in phylogenomics. We reiterate that
there is enormous promise and value in recent implementations and tests of the MSC and look forward
to its increased use and refinement in phylogenomics.

Last updated on 09/22/2016