Biology Reference
In-Depth Information
second part, we will present ongoing research to model more complex
anatomical and developmental data, which could provide the foundation
for a bioinformatic platform for evo-devo studies.
2. The Easy Part: The Evolution
of Gene Sequences
2.1. Rapid Overview of Bioinformatics Involved
The basic task in relating sequence evolution to developmental biology is
finding genes of interest, listing all of their homologs, and determining
their phylogenetic relationships. To study their evolution, we may also be
interested in functional classification as well as in evidence of selective
pressure. For example, a change in selective pressure on some sites may
be evidence of a change in function of the protein.
The first task, identifying genes of interest, might be approached in
two ways: we may start with candidate genes and search for their
homologs, or we may start by determining homologs genome-wide and
use the result of this analysis to select genes of interest. In both cases, a
key step is identifying homologs by sequence similarity. This is a topic
abundantly treated elsewhere, but we need to note here that many genes
of interest in evo-devo are characterized by short conserved domains
which may be difficult to identify in standard scans using, for example,
BLASTP.
31
Transcription factors such as the Hox, bZIP, and bHLH tran-
scription factors are thus best identified by the careful use of hidden
Markov models (e.g. Amoutzias
et al
.
32
), and are often missed in large-
scale scans for homologs. An interesting exception to this is the nuclear
hormone receptor superfamily, whose members can be readily identified
thanks to their ligand-binding domain of approximately 200 amino
acids.
33,34
The topic of genome duplication raises that of the distinction
between orthologs and paralogs.
17,35
The theoretically correct way to
do this is through phylogenetic inference, although numerous alternative
methods have been proposed. We will not treat these methods in detail here,
but note that we systematically use likelihood methods (e.g. PhyML
36
).
While the most common use of phylogenies in such studies is to start
