I am broadly interested in using phylogenetic
trees to understand biology. New methods provide powerful tools for
understanding the processes leading to extant patterns of
morphological, behavioral, biochemical, and species diversity as well
as various types of interactions. I have done empirical work to answer
questions and methods development to create tools to allow us to answer
more questions. I am currently (since November 2007) doing a postdoc at
the National Evolutionary Synthesis
Center (NESCent), where I am working on a database of questions,
methods, and software from comparative biology, genomics, and
paleontology that use a phylogenetic tree and some data to address
evolutionary questions. This database and its associated website,
TreeTapper (development preview at http://treetapper.nescent.org),
will allow users to identify the optimal method and software package to
address a particular question; developers will be able to identify
questions for which there are no methods and methods that are not in
software in order to find where best to put their efforts. I will then
use this database to target holes to fill.
Some of my past and current research work includes: sequencing of
multiple markers for inferring the phylogeny of Myrmecocystus honeypot ants,
developing a method to test for different rates of trait evolution,
investigating morphological and behavioral coevolution in Myrmecocystus ants, developing new
methods for investigating discrete and continuous character
correlation, developing a partitioned likelihood search program,
analyzing supermatrices from Genbank, developing a joint estimator of
species limits and the species tree, developing a method to test for
different rates of gene loss, and more. See the Research page for more.
TreeTapper:
Methods using phylogenetic information
have become
powerful tools for addressing the evolution of morphological,
behavioral, geographical, genetic, biochemical, and physiological
traits. Information on the relevant literature and software is widely
scattered and often segregated into different sub-disciplines, leading
to empiricists choosing to use suboptimal methods or giving up on
certain types of questions and to theoreticians failing to develop
models where needed and occasionally duplicating the work of others. I
am creating a relational database of relevant questions, methods, and
software, and developing a website based on this database (development
site at
http://treetapper.nescent.org;
the final url is
http://www.treetapper.org;
there is also a
progress
blog). Users will eventually be able to search for software to
address a particular question; developers will be able to search for
questions lacking adequate methods and methods not yet implemented in
software. This is different from existing tools, like Felsenstein's
excellent list of
phylogenetic
software, in that it can highlight areas not connected to
software (the white boxes in the diagram on the right), allowing holes
to be filled. There will also
be implementation differences (TreeTapper adds information about links
between authors, how popular various programs or methods are, and a
faster search, but Felsenstein's site covers a broader range of
programs, including those that make, rather than just use, trees).
Problems inspiring this: I want
to build methods, and for that it's good to know where the holes are; I
also got annoyed at papers using what seemed to be less powerful
methods when existing methods, perhaps in a slightly different field,
would have served better, suggesting that finding appropriate methods
is difficult.
Results:
TreeTapper (development in
progress)
Species delimitation:
Delimiting species is a hard problem,
especially because speciation is (often) a gradual process but we
require species to be discrete. There are a variety of conventional
methods for splitting populations into different species (looking for
distinctness in sympatry, looking for traits that would prevent
interbreeding, etc.) and several new "DNA barcoding"-like techniques,
using a tree with branch lengths from a single marker to infer species
boundaries. I have made a new approach to complement existing
approaches: it uses a set of topologies from independent genes and by
looking at the conflict and consistency between the gene trees, infers
the species tree and species boundaries simultaneously, not requiring
any a priori statements about which samples belong in which species.
The first method I created is nonparametric; to my great surprise, it
actually seems to work in many cases (it has been accepted, pending major review, for
Systematic Biology). On the right are ten gene trees
with samples from
Drosophila pseudoobscura, D. persimilis, and
D. ps. bogotana, with name colors
corresponding to species; the program correctly infers the true rooted
species topology, assigning all samples correctly, to return the tree
on the lower right. I am now investigating making a parametric version.
Problem inspiring this:
Brad Shaffer
collecting many gene sequences from what may be multiple species
and lacking a good method for using them to understand species
boundaries.
Results: A new nonparametric
method coded in Brownie for jointly estimating the species tree and
boundary (
in review)
Trait evolution:
One of the reasons I work in this area
of biology is the amazing power using phylogenetic trees gives us for
understanding how evolution occurs. People have used phylogenetic
methods to infer how adaptive radiations occur, how extinct species
attracted mates, how behavior and morphology coevolve, and more. With
collaborators, I made a method to test for different rates of evolution
on different parts of a tree. I've now extended this to allow testing
for correlations between discrete states and continuous rates, discrete
character changes and continuous rates, discrete states and continuous
character means, and other questions. I've also developed a method to
look at branch-specific transition rates to create a test for faster
rates of gene loss in specialist rather than generalist herbivores.
Problems inspiring this:
Peter Wainwright
wanting a measure of disparity correctly taking into account phylogeny;
questions I wanted to address for my ants but for which there were no
appropriate techniques; Lindy McBride having a question about gene loss
in
Drosophila; obvious holes
needing filling
Results: Various methods
implemented in Brownie, publications in
Evolution and
Genetics [the latter a signed
appendix], two more manuscripts nearing submission.
Ant evolution:
Myrmecocystus (Westmael
1838) ants
occur
in arid regions of western North America. They are probably best known
(even to
Darwin)
for their habit of having replete workers: specialized workers with
greatly swollen gasters that store liquids, such as nectar, for times
of scarcity. These ants also do dynamic territory allocation using
ritual combat, have inter- and intraspecific slavery, and, in
some species, do burst foraging, where all the foragers for the day
will leave in one or a few clumps but then forage singly. Different
species, generally divided into different subgenera, will forage at
different times: some species only forage during the hot part of the
desert day, skittering over painfully hot (at least to me) rocks and
picking up dead insects and nectar; some only forage during the cold of
the night, and some only forage at sunset or dawn. Foraging time might
plausibly be correlated with certain morphological traits, like eye
size or leg length; one could also postulate that it might be harder
for the ants to invade the diurnal foraging time period than other time
periods (some ants with other foraging periods die within ten minutes
of exposure during the day). I developed methods to test these ideas.
First, though, I needed a phylogeny, so I sequenced one mitochondrial
and eight nuclear genes for the majority of the species in the group
(some of the missing species have only been collected a handful
of times, sometimes in regions now paved by Los Angeles). To infer the
phylogeny, I wanted to use a mixed model but wanted to avoid the
potential pitfalls of Bayesian methods (see
Yang & Rannala
2005), so I modified
MrBayes
to turn it into a likelihood search program, MrFisher. I also did over
thousand morphological measurements for ecologically-interesting traits
and created interactive online keys (
dichotomous or
multi-entry (
1,
2,
3)),
species
pages, and
range
maps.
Problem inspiring this: Wanting
to understand the evolution of these critters.
Results: Two manuscripts in the
pipeline, new methods & software, new genes for ant phylogeny, a
new phylogeny and measurements, new online resources.
Miscellaneous:
I also address other interesting
problems as they come up. For example, I was part of a
group creating
trees from Genbank using a sparse supermatrix approach. I've also done
phylogenies of
bark
beetles and Dryophthorid weevils (the latter two projects as part
of the Farrell lab as an undergraduate).
PDF Cited by... Related to...
Software for analyzing rate of continuous character evolution and
testing for different rates in different groups. A version in beta
testing allows investigation of correlation of discrete and continuous
traits, many more models for discrete and continuous trait evolution,
species delimitation, and more. There is even a graphical user
interface!
at NESCent
