16
A
ratio test showed that a reflected in the maximum observed
(uncorrected)
likelihood
categories.
estimated
differences across Satumiidae at the
sec-
of invariant
also
sites,
separate category
pairwise
the model's
this is called the GTR+G+I
ond codon
which is 5% for DDC but
likelihood,
(P <0.0005);
by
fit
significantly improved
position, only
1 %for EF-1 œ. Maximum differences at the third
are 55% and
for
and
DDC
model. A
time-reversible model
codon
codon
EF-1
39%
separate general
characters were
and
position
in
which
a,
by
respectively.
the
partitioned
is
In
was also
this
from Satumiidae
position
the
by gene
explored;
When the all-nucleotide
extending
comparisons
Brahmaeidae and
to include the
called
GTR-ssr model.
outgroups
Sphin-
differences at the sec-
pairwise
data set was
ssr model
on the MP
the GTR-
maximum
tree,
a In-likelihood score
gidae,
ond codon
for
optimized
yielded
double
position approximately
11%
at
(to
approxi-
than the
to 2%
for
DDC,
while differences
1.5% lower
versus
GTR+G+I model
EF-la),
mately
Efficient
of the third codon
a
also increase for DDC
for EF-1œ. Our
(to
-18,619).
(-18,342
tree
exploration
entailed
position
remain
in likelihood
but
64%)
rate of an all-taxon data set
analysis
space
analyses
heuristic searches and
unchanged
sepa-
+
(Satumiidae
taxa)
of
(re-)
sequence
parameter
&
et al.
Brahmaeidae +
taxon
64
and a 48-
2000;
(Shultz
optimizations
Regier
Regier
Sphingidae;
Parameters fitted to the ML
were
distance matrices
data set was motivated
2000).
Satumiidae-only
topology
by
differences
used to create corrected
the
that the
pairwise
data sets
possibility
across
high pairwise
in the third
under a mini-
from 1000
mum
codon
families,
bootstrapped
particularly
could mask
model for
evolution
neighbor joining.
Average pairwise sequence
hits,
subsequent analysis by position,
multiple
inference.
undermining
phylogenetic
Both
differences were
show
base
genes frequen-
cies across Satumiidae at all codon
(P
positions
homogeneous
calculated
and
codon
by gene
PAUP*4.0b2. Base
were calculated
by position using
of terminal taxa
values from 0.58 to
further
their
With
the
1.00),
for saturniid
Brahmaeidae and
supporting
phylogenetics.
included,
frequencies
and
codon
by gene
by
position,
suitability
PAUP*4.0b2. The
also
incongruence length
as third
codon
position
Sphingidae
only
using
difference test
the
et al.
for EF -1 œbecomes
(Farris
1995),
implemented
strongly
of com-
in
<
In
0.001).
support
test
PAUP*4.0b2,
(P
partition homogeneity
was used to test for
non-homogeneous
bined
between
of total nucleotides from DDC and
tests reveal no con-
partition homogeneity
analysis
conflicting signal
Relative rates of nucleotide
character
EF-1 a,
partitions.
substitution codon
and
by gene
when
lengths
were
con-
flict in the
from DDC
signals
and
either
EF-1 œ,
inclusion of
by
estimated from
position
tree
=
=
with
without
(P 0.30)
sphingid outgroups.
or
(P 0.11 )
parsimony
strained to the same all-nucleotide-derived
The amount of
was estimated in PAUP*4.0b2 as the difference in
the brahmaeid and
topolo-
nucleotide
change
gy.
synonymous
and
Monophyly, composition,
phylogenetic posi-
discussion
overall tree
derived from total amino acid
from total nucleotide
In
the
results,
tion
lengths
and
when
from
of Saturniinae. -
of our
following
summarized in
3
(separately)
mapped
analysis
change
change
(derived
phylogenetic
Fig.
MP
onto the same
and
we make a heuristic distinction between
4,
topology
nucleotides).
of total
defined as those with
nodes,
(BP) 2:
nodes,
'strongly supported'
and the
80%,
i.e. those with BP <
other,
bootstrap percentages
more
weakly
supported
Results and discussion
EF-
80%. Low BP values can result either from low
numbers characters or from a rela-
and
œ and DDC:
of
supporting
Suitability of
Synonymous
tive abundance of
characters
conflicting
some
(or
pairwise differences,
tests.
non-synonymous changes,
base
-
and
combination).
homogeneity
frequencies,
partition
the all-
the
Within
differences
Likelihood and
of
Satumiidae,
about 92% total
of which resides at the third codon
synonymous
of character
parsimony analyses
data 64 taxa
set for
for
nucleotide
account
variation,
strongly support
of Satumiidae from the other two fam-
most
position
between
separation
ilies
Satumiidae,
=
and is
DDC
Within
are
there is
100%;
of
(BP
the two
3).
Fig.
approximately equally partitioned
sampled
and
EF-l œ.
contrast,
By
are more than threefold
non-synonymous
greater
genera Ceratocampinae
for DDC
However,
91%).
any
sampled. Similarly
strongly
no
changes
grouped
support
three subfamilies
than EF-l œ
also
et al.
for
the
(see
non-synonymous change
1998).
strong
among
Regier
Higher
grouping
weak resolu-
rates of
for DDC are also