83
5. The root of a phylogenetic tree is the node corresponding to
the ancestor of all sequences in the tree. Rooted trees show the
position of the root and the direction of evolution, whereas the
unrooted trees produced by the maximum-likelihood method
contain no information about the direction of evolution. This
problem can be circumvented by the inclusion of outgroup
sequences, which are known to be more distantly related to the
sequences of interest (ingroup sequences), in the phylogenetic
analysis. The root of the tree will be located somewhere on the
branch joining the ingroup and outgroup sequences.
6. Various alignment programs with different specialities are available. In addition to MUSCLE, these include MAFFT [33],
T-COFFEE [34], and PRANK [35]. In particular, PRANK
may be useful for APR, since its phylogeny-aware algorithm is
able to distinguish independent insertion and deletion events
that would result in erroneous inferences of homology in conventional multiple sequence alignment programs [36].
7. Programs for automatic curation of multiple sequence alignments are available (such as Gblocks [37]), but these programs
may be too zealous in removing poorly aligned regions of the
alignment for this particular application. Although insertions
unique to individual sequences or a small cluster of sequences
can be safely removed, care should be taken not to delete columns that might contain residues present in the ancestral proteins of interest. Since phylogenetic information is needed to
make this judgement, the alignment should be edited conservatively before the reconstruction of ancestral sequences.
Artifactual insertions in the ancestral sequences can be
removed afterward, once the phylogeny is known.
8. More recent implementations of maximum-likelihood methods may give superior performance in terms of computational
speed and/or accuracy. Alternative software options include
IQ-TREE [38] for model selection, IQ-TREE or RAxML
[39] for tree inference, and FastML [40] for reconstruction of
ancestral sequences.
9. The assumption implicit in this choice is that the BIONJ tree
is a good enough approximation to the maximum-likelihood
tree that the substitution model of best fit will be identical for
both trees. It is more accurate to use the maximum-likelihood
tree as the base tree for the likelihood calculations, but the
computation will be much slower.
10. If these settings are used, the tree search will be deterministic,
that is, the same alignment will always produce the same tree.
This can be problematic, since the tree search may stall at a
locally optimal tree and fail to converge on globally optimal
tree (the true maximum-likelihood tree). It is therefore good
Improving FRET Sensors by Ancestral Gene Resurrection
5. The root of a phylogenetic tree is the node corresponding to
the ancestor of all sequences in the tree. Rooted trees show the
position of the root and the direction of evolution, whereas the
unrooted trees produced by the maximum-likelihood method
contain no information about the direction of evolution. This
problem can be circumvented by the inclusion of outgroup
sequences, which are known to be more distantly related to the
sequences of interest (ingroup sequences), in the phylogenetic
analysis. The root of the tree will be located somewhere on the
branch joining the ingroup and outgroup sequences.
6. Various alignment programs with different specialities are available. In addition to MUSCLE, these include MAFFT [33],
T-COFFEE [34], and PRANK [35]. In particular, PRANK
may be useful for APR, since its phylogeny-aware algorithm is
able to distinguish independent insertion and deletion events
that would result in erroneous inferences of homology in conventional multiple sequence alignment programs [36].
7. Programs for automatic curation of multiple sequence alignments are available (such as Gblocks [37]), but these programs
may be too zealous in removing poorly aligned regions of the
alignment for this particular application. Although insertions
unique to individual sequences or a small cluster of sequences
can be safely removed, care should be taken not to delete columns that might contain residues present in the ancestral proteins of interest. Since phylogenetic information is needed to
make this judgement, the alignment should be edited conservatively before the reconstruction of ancestral sequences.
Artifactual insertions in the ancestral sequences can be
removed afterward, once the phylogeny is known.
8. More recent implementations of maximum-likelihood methods may give superior performance in terms of computational
speed and/or accuracy. Alternative software options include
IQ-TREE [38] for model selection, IQ-TREE or RAxML
[39] for tree inference, and FastML [40] for reconstruction of
ancestral sequences.
9. The assumption implicit in this choice is that the BIONJ tree
is a good enough approximation to the maximum-likelihood
tree that the substitution model of best fit will be identical for
both trees. It is more accurate to use the maximum-likelihood
tree as the base tree for the likelihood calculations, but the
computation will be much slower.
10. If these settings are used, the tree search will be deterministic,
that is, the same alignment will always produce the same tree.
This can be problematic, since the tree search may stall at a
locally optimal tree and fail to converge on globally optimal
tree (the true maximum-likelihood tree). It is therefore good
Improving FRET Sensors by Ancestral Gene Resurrection