Comparative Genomics-Basic and Applied Research.pdf

34 Comparative Genomics
quickly but producing a good one is time consuming as most optimization criteria
are nondeterministic polynomial-time hard (NP-hard). Reliability, the ability of a
reconstruction method to return accurate answers on entirely new data sets rather
than just those on which it has been tested (and often developed), remains largely
unexplored; while systematists are accustomed to getting so-called bootstrap
scores for their tree edges or estimates of distributions of trees from their Markov
chain Monte Carlo (MCMC) methods, the predictive value of the reconstruction
methods and the significance on any given sample data set of these quality measures
remain mostly unknown.
Surprises have been encountered time after time as the scale of reconstruction
increased; thus, current methods, even if reliably accurate within their current ranges
(something we do not know), are not likely to remain so as we move to larger scales.
3.1.5 RECONSTRUCTING THE TREE OF LIFE
Many biologists have been calling for some time for a community effort to attempt
the reconstruction of the tree of life, the phylogeny of all organisms on this planet.
Such an endeavor naturally has no end since evolution is an ongoing process and is
not particularly well defined since thousands of organisms become extinct every
year, if not every day. The scale is truly daunting: While we have methods that can
reconstruct phylogenies for up to a thousand leaves (and scale poorly beyond that),
there are well over a million described species of organisms, and estimates of the
existing number vary from ten million to several hundred millions. Finally, it is not
clear that we need a single giant phylogeny; many of the branches of this phylogeny
are well identified and broadly accepted and so could be investigated mostly independently
of all others. Yet, the tree of life should hold a special place in the heart
of every human: It describes the wonderful diversity of life on this planet, helps us
understand where we humans come from and what is our place within the larger
scheme of life, and most importantly, gives us a basis to understand where we are all
heading. The project to reconstruct this phylogeny also motivates the community to
revisit many aspects of phylogenetic analysis, particularly those that have to do with
scaling and reliability. After all, there is only one tree of life for this planet, so there
will not soon be a chance to compare our reconstruction with one done for another
tree of life elsewhere.
In the United States, the National Science Foundation initiated the Assembling
the Tree of Life program that has funded, to date, well over 30 groups collecting,
filtering, and analyzing data on all branches of the tree. Through another program, it
has also enabled the Cyberinfrastructure for Phylogenetic Research (CIPRES) project
(www.phylo.org), with the aim to develop the informatics infrastructure (software
framework, databases, analysis modules, workflow, and hardware platform)
necessary to attack the computational problems that the community will face in
attempting a reconstruction of the tree of life. Many other research groups throughout
the world are working on the tree of life in some form. The resulting surge of
interest in large-scale phylogenetic reconstruction from combinatorialists, statisticians,
algorithm designers, high-performance computing specialists, and of course,
biologists and biomedical researchers has begun to yield spectacular results.