[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Molecules in Theropod Phylogen [was: Re: Theropod phylogenies]

HP Holtz wrote (quoting HP Forrest):
> > As I understand it, sorting out the relationships of modern birds based
> > morphology alone is tricky, and the advent of firstly molecular data,
> > more recently DNA analysis has led to some extensive revision of their
> > relationships.
> Well, to be fair, DNA analyses will often give wildly different topologies
> depending on the particular genes and the particular subset of taxa

    WARNING: the following post contains gross overgeneralizations. Deal.

    To support HP Holtz's point: there are a number of issues with molecular
data that contradict the popular assumption that molecular data is
inherently superior to morphology. Chief among them, in my opinion, is the
potential for any character (e.g., nucleotide position) to tranform to any
of the possible character states. For morphological characters, each
character state represents a novel morphology that is (except in the case of
reversals) entirely new to that lineage. In a gene sequence, each character
state is one of the "built-in" options, and is fair game as a potential
transformation. Compounded upon this is the EXPECTATION that many characters
WILL change over some period of time. Although morphology does change, and
can even erase the record of previous change, it is not expected to. Many
nucleotide positions (e.g., the "third position" in a codon) are under very
little selection pressure, and replication errors can move to fixation
within populations reltively quickly (e.g., within species or among sister
species). Replication errors are expected, therefore change is expected.
    This is not to say that molecular data are "bad," just that they
"behave" a little differently from morphological studies. In gel-jock terms,
you can choose the "wrong gene" for the relationships you are trying to
explore. Because selection pressures, structural attributes of the gene or
gene product, and other parameters differ among genes and even among
nucleotide positions WITHIN genes, the process of sequence evolution
proceeds at different rates within and among genes. If the sequence in
question evolved "too quickly," change may have effectively wiped out the
record of the relationships you are trying to find. For example, cytochrome
B sequences appear to be misinformative with regard to the ordinal
relationships of birds. In some cases, it may wipe out one or more
heirarchical "levels" of the phylogenetic signal, but not others, e.g.,
cytochrome B sequences that tend tend to give expected relationships at the
"subfamily" or "family" level for turtles, but not a the "suprafamilial"
level; cyt B sequences ARE informative within bird "families." If the
sequence is evolving too slowly, a few homoplastic changes can override the
true phylogenetic signal, as is probably the case in several studies using
the 18S nuclear ribosomal gene to elucidate vertbrate phylogeny, but which
have grouped birds and mammals. Many of the commonly applied "support"
statistics cannot tell you you have this problem, because they only test the
fit of the tree to the data, not the fit of the data to reality.
    Morphology can be thought of as a variable-rate "gene" that generally
transforms at a rate such that it gives good resolution at the levels
traditionally termed "classes" to about the "genus" or "family" level. Below
this level, morphologic change may not keep pace with lineage splitting (one
possible explanation for "cryptic species"), and above this level, change
often erases the evidence of relationships, or adds characters that have no
homologs in other taxa (effectively an "insertion" in the "sequence"). Thus,
we still have questions about how the "phyla" relate to one another, but we
have more confidence in class, order, and family relationships. IMHO,
morphology may be misleading as to the relationships of closely related
    For both types of data, phylogeny is most easily recovered when there is
a moderate rate of evolution, and a relatively "even" tree, with the amount
of evolution between nodes being comparable to the amount of divergence
between the descended branches. In cases where there has been a rapid
radiation ("short internodes" in gelspeak), possibly coupled with a high
degree of change within lineages after divergence ("long branches"), neither
method provides particularly convincing results. Crown birds and placental
mammals (and possibly Neobatrachian frogs, salamanders and non-iguanian
squamates) apparrently fit this description. The hope is that molecular data
will get around issues of morphological homoplaisy and morphological "long
branches," but then the issue is finding the "right gene." Morphology has a
partial advantage in these situations, because fossils can "break up long
branches," but it still suffers in the case of short divergence times at the
base of a radiation.
    Also, there are methodological issues with both types of data, such as
character coding, which in molecular studies is paralleled by sequence
alingment (sequences are not pre-numbered to tell you which nucleotide in
sequence A corresponds to which nucleotide in sequence B), outgrouping (the
outgroup branch is often the longest one in the tree), and taxon selection
(as noted by HP Holtz), all of which can compound other problems to give a
widely disparate result. Despite the arrogance of some proponents of both
methods, neither is inherently superior, and many prominent systematists
accept both classes of data, along with others (behaviour, karyology).
Indeed, in many recent studies, the standard by which new molecular methods
and/ or new trees from molecular data have been evaluated is, in fact, the
morphological tree.

    So, the upshot is: do NOT assume that one source of data is inherently
superior to the other. As noted by HP Holtz, the trees for MANY extant
groups are still quite preliminary. As far as I can tell, we can consider
the verdict to be still "out" on at LEAST "higher" teleost fish,
salamanders, neobtrachian frogs, scleroglossan lizards (including
non-scolecophidian snakes), placentals, neoavian birds (especially
passerines!), and maybe Crocodylus as well, among Osteichthyes. This isn't
meant to be a comprehensive list, by the way. These are groups that have
been subjected to few comprehensive morphological and/ or molecular studies
in a modern (cladistic) framework. Also, in some cases (e.g., birds, frogs),
the time differential between the first comprehensive morphological and
molecular analyses for the group is FAR too short (e.g., a decade or so at
most) to really say that molecular data have overturned ANYTHING. A prudent
and forward-looking systematist, IMHO, uses all of the data available to him
(either in combination or in contrast) to evaluate results, and questions
the methods used to arrive at the tree rather than simply accepting one tree
over others due to personal bias.