4 Phylogeny of Animals
131
lineage has a high chance to be lost in another one (Zheng et al. 2007). Initially,
support was recovered for Ecdysozoa (Roy and Gilbert 2005) but this result was subsequently criticized for not having sufficiently accounted for parallel intron losses,
and contradictory analyses recovered support for Coelomata (Zheng et al. 2007).
However, it has been suggested that this latter result was prone to the same rooting problems as the analyses of rare amino acid changes. Subsequent incorporation
of the intron-rich cnidarian N. vectensis provided support for the Ecdysozoa again
(Roy and Irimia 2008). Thus, although the importance of taxonomic sampling has
been extensively discussed for conventional, sequence-based molecular phylogenies, there is clearly a need for similar attention to be paid to this problem for new
types of phylogenetic evidence such as rare genomic changes (Rokas and Holland
2000b).
4.4.2 Is It Actually Possible to Decipher Animal Relationships?
A more general concern about the efficiency of phylogenomics has been raised
by Rokas et al. (2003), who question its ability to fully resolve metazoan relationships (Rokas et al. 2005). To explain their failure to recover several classical
metazoan clades using a 17 taxa and 50 gene dataset, the authors proposed that
metazoan radiation was “compressed in time”, which means that the time of divergence between animals lineages would have been too short for the deployment of
an accurate phylogenetic signal. Despite its stimulating ideas, this work was contradicted by numerous other phylogenomic studies (Marlétaz et al. 2006, Matus et al.
2006, Philippe et al. 2005b). The problem of the lack of resolution observed by
Rokas et al. (2005) was specifically addressed in subsequent work (Baurain et al.
2007) and two primary causes were identified: inappropriate taxon sampling and
misleading evolutionary models. First, no effort was made to select slow evolving
species within the diverging groups such as nematodes, although ecdysozoans were
for example easily recovered when the nematode Xiphinema was employed instead
of Caenorhabditis (Baurain et al. 2007). Also, it has been shown that the resolution
of the tree may be improved by the use of better amino acid substitution models
such as the CAT model or of better tree search algorithms such as SPR (subtree
pruning and rebranching) (Hordijk and Gascuel 2005, Lartillot et al. 2007). Finally,
the poor quality of raw sequence data could also hinder node resolution. Rokas et al.
(2005) retrieved most of their sequences through PCR amplifications performed on
conserved domains, thereby excluding the most variable and phylogenetically informative regions of the genes they analysed. Other phylogenomic studies of animal
relationships have been mainly based on EST data that yield larger gene fragments.
The accuracy of phylogenomic reconstruction could be greatly enhanced by the
selection of slowly diverging species, a strategy closely related to that proposed by
Aguinaldo et al. for SSU-based phylogeny (Aguinaldo et al. 1997). However, in
the multigene approach, it is not clear whether a single species possesses the least
diverging copies of all marker genes. To overcome this problem, Marlétaz et al.
(2006) have proposed the use of a new composite taxon strategy that selects the
131
lineage has a high chance to be lost in another one (Zheng et al. 2007). Initially,
support was recovered for Ecdysozoa (Roy and Gilbert 2005) but this result was subsequently criticized for not having sufficiently accounted for parallel intron losses,
and contradictory analyses recovered support for Coelomata (Zheng et al. 2007).
However, it has been suggested that this latter result was prone to the same rooting problems as the analyses of rare amino acid changes. Subsequent incorporation
of the intron-rich cnidarian N. vectensis provided support for the Ecdysozoa again
(Roy and Irimia 2008). Thus, although the importance of taxonomic sampling has
been extensively discussed for conventional, sequence-based molecular phylogenies, there is clearly a need for similar attention to be paid to this problem for new
types of phylogenetic evidence such as rare genomic changes (Rokas and Holland
2000b).
4.4.2 Is It Actually Possible to Decipher Animal Relationships?
A more general concern about the efficiency of phylogenomics has been raised
by Rokas et al. (2003), who question its ability to fully resolve metazoan relationships (Rokas et al. 2005). To explain their failure to recover several classical
metazoan clades using a 17 taxa and 50 gene dataset, the authors proposed that
metazoan radiation was “compressed in time”, which means that the time of divergence between animals lineages would have been too short for the deployment of
an accurate phylogenetic signal. Despite its stimulating ideas, this work was contradicted by numerous other phylogenomic studies (Marlétaz et al. 2006, Matus et al.
2006, Philippe et al. 2005b). The problem of the lack of resolution observed by
Rokas et al. (2005) was specifically addressed in subsequent work (Baurain et al.
2007) and two primary causes were identified: inappropriate taxon sampling and
misleading evolutionary models. First, no effort was made to select slow evolving
species within the diverging groups such as nematodes, although ecdysozoans were
for example easily recovered when the nematode Xiphinema was employed instead
of Caenorhabditis (Baurain et al. 2007). Also, it has been shown that the resolution
of the tree may be improved by the use of better amino acid substitution models
such as the CAT model or of better tree search algorithms such as SPR (subtree
pruning and rebranching) (Hordijk and Gascuel 2005, Lartillot et al. 2007). Finally,
the poor quality of raw sequence data could also hinder node resolution. Rokas et al.
(2005) retrieved most of their sequences through PCR amplifications performed on
conserved domains, thereby excluding the most variable and phylogenetically informative regions of the genes they analysed. Other phylogenomic studies of animal
relationships have been mainly based on EST data that yield larger gene fragments.
The accuracy of phylogenomic reconstruction could be greatly enhanced by the
selection of slowly diverging species, a strategy closely related to that proposed by
Aguinaldo et al. for SSU-based phylogeny (Aguinaldo et al. 1997). However, in
the multigene approach, it is not clear whether a single species possesses the least
diverging copies of all marker genes. To overcome this problem, Marlétaz et al.
(2006) have proposed the use of a new composite taxon strategy that selects the