[转帖]nature上最新的系统生物学文章

王不留行

Vol. 438|22/29 December 2005
COMMENTARY
1079
Barriers to progress in systems biology
For the past half-century, biologists have been uncovering details of countless molecular events. Linking these
data to dynamic models requires new software and data standards, argue Marvin Cassman and his colleagues.
The field of systems biology is lurching
forwards, propelled by a mixture of
faith, hope and even charity. But if it is
to become a true discipline, several problems
with core infrastructure (data and software)
need to be addressed. In our view, they are too
critical to be left to ad hoc developments by
individual laboratories.
Systems biology has been defined in many
ways, but has at its root the use of modelling
and simulation, combined with experiment, to
explore network behaviour in biological
systems — in particular their dynamic nature.
The need to integrate the profusion of
molecular data into a systems approach has
stimulated growth in this area over the past
five or so years, as worldwide investments
in the field have increased. However, this
early enthusiasm will need
to overcome several barriers
to development.
A recent survey carried
out by these authors —
conducted by the World
Technology Evaluation
Center (WTEC) in Baltimore,
Maryland, and
funded by seven US agencies
— compared the activities of system biologists
in the United States, Europe and Japan1.
The survey reveals that work on quantitative
or predictive mathematical modelling that is
truly integrated with experimentation is only
just beginning. Progress is limited, therefore,
and major contributions to biological understanding
are few. The survey concludes that
the absence of a suitable infrastructure for systems
biology, particularly for data and software
standardization, is a major impediment
to further progress.
Come together
The WTEC survey confirmed that vital software
is being developed at many locations
worldwide. But these endeavours are highly
localized, resulting in duplicated goals and
approaches. Tellingly, one Japanese group
called their software YAGNS, for ‘yet another
gene network simulator’. There are many reasons
for this cottage industry: the need to
accommodate local data; the requirements of
collaborators to visualize data; and limited
knowledge of what is already available. In
general, however, it is a terrible waste of time,
money and effort. Most software remains inaccessible
to external users, even when the
developers are willing to release it, because
supporting documentation is so poor.
For software developers and skilled users
these problems are not insurmountable. But
sharing of the benefits of systems biology
more widely will occur only when working
biologists, who are not themselves trained to
develop and modify such software, can
manipulate and use these techniques.
Unfortunately, the translation of systems
biology into a broader approach is complicated
by the innumeracy of many biologists.
Some modicum of mathematical training
will be required, reversing the trend of the
past 30 years, during which biology has
become a discipline for
people who want to do
science without learning
mathematics.
A reasonable set of
expectations is that different
pieces of shared software
should work together
seamlessly, be transparent
to the user, and be
sufficiently documented so that they can be
modified to suit different circumstances.
Funding agencies would be unwise to support
software development without also investing
in the infrastructure needed to preserve and
enhance the results. One way to do this would
be to create a central organization that would
serve both as a software repository and as a
mechanism for validating and documenting
each program, including standardizing of the
data input/output formats.
As with centralized databases, having a
shared resource with appropriate softwareengineering
standards should encourage users
to reconfigure the most useful tools for increasingly
sophisticated analysis. A group sponsored
by the US Defense Advanced Research Projects
Agency, and involving one of us (M.C.), has
developed a proposal for such a resource2. This
repository would serve as a central coordinator
to help develop uniform standards, to direct
users to appropriate online resources, and to
identify — through user feedback — problems
with the software. The repository should be
organized through consultation with the
community, and will require the support of an
international consortium of funding agencies.
Diverse data
The problems with software diversity are
mirrored by the diversity of ways that data are
collected, annotated and stored. Such issues are
even worse than those faced by the DNAsequencing
community, because experimental
data in systems biology is highly context
dependent. For data to be useful outside the
laboratory in which they were generated, they
must be standardized, presented using a uniform
and systematic vocabulary, and annotated
so that the specific cell type, growing conditions
and measurements made — from metaboliteand
messenger-RNA-profiling to kinetics and
thermodynamics — are reproducible.
Easy access to data and software is not a
luxury, it is essential when results undergo
peer review and publication. For the scientific
community to evaluate the increasingly
complex data types, the increasingly sophisticated
analysis tools, and the increasingly
incomplete papers (that cannot include all
information because of the very complexity of
the experiments and tools), it is vital that it has
access to the source data and methods used.
Dealing with these complex infrastructure
issues will require a focused effort by
researchers and funding agencies. We propose
that the annual International Conferences on
Systems Biology would be an appropriate venue
for initial discussions. Whatever the occasion, it
must be done soon. ■
Marvin Cassman lives in San Francisco,
California, USA.
Co-authors are Adam Arkin of the Bioengineering
Department, University of California, Berkeley;
Fumiaki Katagiri of the Department of Plant
Biology, University of Minnesota, St Paul;
Douglas Lauffenburger of the Biological
Engineering Division, Massachusetts Institute of
Technology, Cambridge; Frank J. Doyle III of the
Department of Chemical Engineering, University
of California, Santa Barbara; and Cynthia L. Stokes
who is at Entelos, Foster City, California.
1. Cassman, M. et al. Assessment of International Research and
Development in Systems Biology (Springer, in the press)
www.wtec.org/sysbio
2. Cassman, M., Sztipanovits, J., Lincoln, P. & Shastry, S. S.
Proposal for a Software Infrastructure in Systems Biology
www.csl.sri.com/users/lincoln/SystemsBiology/SI.doc
“During the past 30 years
biology has become a
discipline for people who
want to do science without
learning mathematics.”
Nature PublishingGrou© 2005 p

王不留行

[这个贴子最后由王不留行在 2006/01/01 09:55pm 第 1 次编辑]

这篇文章是两天前nature发表的，我试着翻译翻译，英语水平有限，见谅
题目：系统生物学发展的障碍

王不留行

For the past half-century, biologists have been uncovering details of countless molecular events. Linking these
data to dynamic models requires new software and data standards, argue Marvin Cassman and his colleagues
引言：马文和他的研究小组认为：在过去的半个世纪，生物学家们已经发现了数不清的分子活动的细节，把这些（细节的）资料联系起来整理成动态的模型需要新的软件和数据标准。

王不留行

这篇是cell上的
1174 Cell 123, December 29, 2005 ©2005 Elsevier Inc.
a given problem more holistically. Most
visionaries of the past are forgotten
because their grand ideas and books
became useless once the pedestrian
way of experimental science revealed
their incompatibilities with the facts of
nature. Science remains the art of the
solvable. Traveling the systems road,
we must constantly ask ourselves how
appropriate the big picture is and how
adequate the systems approach is to
the level of the question we are trying to
answer. The fundamentally new characteristic
of systems biology is its way
of thinking, rather than its way of doing.
Systems thinking realizes that the phenotype
of a system (from the shape of
a cell to an evolutionary stable strategy)
is the emergent property of the interactions
among all of the components
of this system. Thus, it is neither the
scale of the system nor the particular
approach used to arrive at a list of its
functional components that defines a
systems approach. In fact, perhaps
paradoxically, for research driven by
this concept to succeed, it may be
necessary first to isolate a reduced
system to provide an experimentally
testable hypothesis. For example, to
understand the molecular changes
that occur in a cell upon binding of a
ligand to its receptor, most quantitative
biologists largely query well-defined in
vitro cell culture systems, which do not
necessarily reflect the in vivo responses
of a complex developing system. Thus,
for the time being, the practical (as
opposed to the conceptual) translation
of systems biology is much better
referred to as large-scale reductionism.
A further complication is that every
system can be described at numerous
levels, but only very few of these
are relevant to a useful understanding
of the system. To give an example, the
early universe, a car engine, and a boa
constrictor are all products of quantum
interactions of subatomic particles. Yet
a quantum description of these interactions
is only useful for one of the three
systems: it can neither tell us if the
engine is working nor what the snake
had for lunch. Richard Dawkins refers
to this necessary feature of scientific
inquiry as “hierarchical reductionism”
(Dawkins, 1986). So, although largescale
measurements are imperative
for a comprehensive description of the
system, the level at which both measurements
and integration occur must
vary depending on the system being
studied and the question being asked.
Our third consideration questions
the assumption that if systems biology
is holistic, then genetics is reductionist.
Let us first have a closer look at the
“omics” approach. It is now possible to
measure, with increasing precision and
in some cases in real time, the molecular
constituents of a system and their
variations across a series of dynamic
phenotypic changes. These measurements
are collectively referred to as
“omics” (as in genomics, transcriptomics,
proteomics, lipidomics, and so
on). But not every “omics” experiment
is systems biology. It depends on the
question. If the purpose of a microarray
experiment, for example, is to identify
a few target genes for a transcription
factor and then validate the “most
promising candidates,” then this is not
systems biology. If, on the other hand,
the purpose is to describe the global
transcriptional response of the cell to
changes in the level, localization, or
sequence of the transcription factor
and then ask how the new molecular
conditions created in the cell interact
to produce the phenotype, then that
is systems biology. Thus, the tools put
constraints on the task at hand, but
they do not define it. So, what about
the genetic approach?
We argue that the assumption that
genetics, and especially forward genetics,
is a reductionist approach is simply
erroneous. Like a microarray experiment,
a genetic screen is not itself
reductionist or holistic. It is the use of
the genetic toolbox that defines its outcome.
It seems that, by mistaking the
“omics” wave for the systems approach
itself, we are forgetting some of the
most influential systems approaches
of the past: when Christiane Nusslein-
Volhard and Eric Wieschaus (Nusslein-
Volhard and Wieschaus, 1980) targeted
the whole Drosophila genome
using random mutagenesis to unravel
the riddle of embryonic pattern formation,
they were doing systems biology.
Other classical examples include the
Drosophila screens of Seymour Benzer
(for example Hotta and Benzer,
1972) and the C. elegans screens of
Sydney Brenner (for example, Hodgkin
and Brenner, 1977). How conceptually
different is a genome-wide forward
genetic screen from genome-wide
RNAi screens (reviewed in Friedman
and Perrimon, 2004)? Today, mouse
RNAi screens and proteomics measurements
can only be done in vitro. As
such, is a forward genetic screen for
behavioral defects in the living mouse
not at least as much, if not more, relevant
to systems biology? In general, a
genetic screen addresses the following
questions: what is the total number of
components required to build a given
phenotype (system) and what is the
contribution of each of these components
to the phenotype? To answer
these questions, genetics assumes
(correctly) that perturbation of these
components should result in some
change in the expression of the phenotype.
Furthermore, our ever-increasing
capacity to visualize and quantify
subtle and dynamic phenotypes—from
cell shape to behavior—in live animals
means future genetic screens will provide
an unprecedented wealth of physiologically
relevant information. Genetics
is not only compatible with systems
biology, it is a corner stone of any useful
form of it. But if all it takes to remain
fashionable is a fresh label, “Forward
Genetomics” might do nicely.
References
Dawkins, R. (1986). The Blind Watchmaker, 1st
Edition (New York: Norton & Company).
Friedman, A., and Perrimon, N. (2004). Curr.
Opin. Genet. Dev. 14, 470–476.
Gregan, J., Rabitsch, P.K., Sakem, B., Csutak,
O., Latypov, V., Lehmann, E., Kohli, J., and Nasmyth,
K. (2005). Curr. Biol. 15, 1663–1669.
Gunsalus, K.C., Ge, H., Schetter, A.J., Goldberg,
D.S., Han, J.D., Hao, T., Berriz, G.F., Bertin, N.,
Huang, J., Chuang, L.S., et al. (2005). Nature
436, 861–865.
Hodgkin, J.A., and Brenner, S. (1977). Genetics
86, 275–287.
Hotta, Y., and Benzer, S. (1972). Nature 240,
527–535.
Kirschner, M.W. (2005). Cell 121, 503–504.
Liu, E.T. (2005). Cell 121, 505–506.
Nagy, A., Perrimon, N., Sandmeyer, S., and
Plasterk, R. (2003). Nat. Genet. Suppl. 33,
276–284.
Nusslein-Volhard, C., and Wieschaus, E. (1980).
Nature 287, 795–801.