P E R S P E C T I V E S
41. Sagar, M. A., Bullivant, D. P., Mallinson, G. D.,
Hunter, P. J. & Hunter, I. W. A virtual environment and
model of the eye for surgical simulation. Compute.
Graph. (ACM) 205–212 (Siggraph, ACM, Addison
Wesley, Ontario, 1994).
Acknowledgements
The authors gratefully acknowledge the contributions from mem-
bers of the Bioengineering Institute at the University of Auckland,
New Zealand. P.J.H. acknowledges the support of the New
Zealand Foundation for Research, Science and Technology, the
New Zealand Health Research Council and the Wellcome Trust.
He is also grateful for the discussions on the Physiome Project,
over many years, with D. Noble (Oxford University, UK),
J. Bassingthwaighte (University of Washington in Seattle, USA)
and A. McCulloch (University of California at San Diego, USA).
Online links
FURTHER INFORMATION
AnatML: http://www.physiome.org.nz/sites/physiome/
anatml/pages/index.html
BioPSE: http://software.sci.utah.edu/biopse.html
CellML: http://www.cellml.org
CMISS:
http://www.bioeng.auckland.ac.nz/cmiss/cmiss.php
Continuity: http://cmrg.ucsd.edu/modelling/software
E-Cell project: http://www.e-cell.org
Gene Ontology Consortium: http://www.geneontology.org
Gepasi: http://www.gepasi.org
Global open biological ontologies (GOBO):
http://www.geneontology.org/doc/gobo.html
The Bioengineering Institute:
http://www.bioeng.auckland.ac.nz
The IUPS Physiome Project:
http://www.bioeng.auckland.ac.nz/physiome/physiome.php
The Microcirculation Physiome Project:
http://www.bme.jhu.edu/news/microphys
The National Resource for Cell Analysis and Modeling:
http://www.nrcam.uchc.edu
The World Wide Web Consortium: http://www.w3c.org
Systems Biology Workbench: http://www.sbw-sbml.org
Virtual Cell:
http://www.nrcam.uchc.edu/vcell_development/vcell_dev.html
Access to this interactive links box is free online.
Peter J. Hunter is at the Bioengineering Institute,
University of Auckland, New Zealand.
Thomas K. Borg is at the Department of
Developmental Biology and Anatomy,
University of South Carolina, USA.
Correspondence to P.J.H.
e-mail: p.hunter@auckland.ac.nz
doi:10.1038/nrm1054
1. International Human Genome Mapping Consortium.
A physical map of the human genome. Nature 409,
934–941 (2001).
2. Venter, C. et al. The sequence of the human genome.
Science 291, 1304–1351 (2001).
3. Noble, D. Biological computation. Encyclopedia of Life
Sciences [online], (DOI 10.1038/npg.els.0003433),
<http://www.els/net> (2002).
4. Noble, D. The rise of computational biology. Nature
Rev. Mol. Cell Biol. 3, 460–463 (2002).
5. Kitano, H. Systems biology: a brief overview. Science
295,1662–1664 (2002).
6. Kitano, H. Computational systems biology. Nature 420,
206–210 (2002).
7. Rao, C. V., Wolf, D. M. & Arkin, A. P. Control,
exploitation and tolerance of intracellular noise. Nature
420, 231–237 (2002).
8. Goldbeter, A. Computational approaches to cellular
rhythms. Nature 420, 238–245 (2002).
9. Edelstein-Keshet, L. Mathematical Models in Biology
(Random House, New York, 1988).
10. Keener, J. & Sneyd, J. Mathematical Physiology
(Springer, New York, 1998).
11. Bower, J. M. & Bolouri, H. (eds). Computational
Modeling of Genetic and Biochemical Networks
(MIT Press, Cambridge, Massachusetts,
2001).
12. Fall, C. P., Marland, E. S., Wagner, J. M. & Tyson, J. J.
Computational Cell Biology (Springer, New York,
2002).
13. Christie, G. R., Bullivant, D. P., Blackett, S. A. &
Hunter, P. J. Modelling and visualising the heart.
Computing. Vis. Sci. 4, 227–235 (2002).
14. Kohl, P., Noble, D. & Hunter, P. J. (eds). The integrated
heart: modelling cardiac structure and function.
Phil. Trans. R. Soc. A 359 (2001).
15. Smith, N. P. et al. Mathematical modelling of the heart:
cell to organ. Chaos, Solitons Fractals 13, 1613–1621
(2001).
16. Smith, N. P., Pullan, A. J. & Hunter, P. J. An anatomically
based model of transient coronary blood flow
in the heart. SIAM J. Appl. Math. 62, 990 –1018
(2001).
17. Barth, T. J., Chan, T. & Haimes, R. (eds). Multiscale and
Multiresolution Methods. Lecture Notes in
Computational Science and Engineering (Springer,
Berlin, 2002).
18. Antzelovitch, C. et al. Influence of transmural gradients
on the electrophysiology and pharmacology of
ventricular myocardium. Cellular basis for the Brugada
and long-QT syndromes. Phil. Trans. R. Soc. A 359,
1201–1216 (2001).
19. Noble, D. Unraveling the genetics and mechanisms of
cardiac arrhythmia. Proc. Natl Acad. Sci. USA 99,
5755–5756 (2002).
20. Hedley, W. J. H., Nelson, M. R., Bullivant, D. P. &
Nielsen, P. F. A short introduction to CellML.
Phil. Trans. R. Soc. A 359, 1073–1089 (2001).
21. Bock, G. R. & Goode, J. A. (eds). The limits of
reductionism. Novartis Foundation Symp. 213
(John Wiley, London, 1998).
22. Bock, G. R. & Goode, J. A. (eds). Complexity in
biological information processing. Novartis Found.
Symp. 239 (John Wiley, London, 2001).
23. Bock, G. & Goode, J. (eds) In silico simulation of
biological processes. Novartis Found. Symp. 247
(John Wiley, London, 2002).
24. Kitano, H. in Foundations of Systems Biology
(ed. Kitano, H) 1–36 (MIT Press, Cambridge,
Massachusetts, 2002).
25. Kohl, P., Noble, D., Winslow, R. L. & Hunter, P. J.
Computational modelling of biological systems: tools
and visions. Phil. Trans. R. Soc. A 358, 579–610
(2000).
26. Bassingthwaighte, J. B. Strategies for the Physiome
Project. Ann. Biomed. Eng. 28, 1043–1058
(2000).
27. LeGrice, I. J., Hunter, P. J. & Smaill, B. H. Laminar
structure of the heart: a mathematical model.
J. Physiol. 272, H2466–H2476 (1997).
28. LeGrice, I. J., Hunter, P. J., Young, A. A. & Smaill, B. H.
The architecture of the heart: a data-based model.
Phil. Trans. R. Soc. A 359, 1217–1232 (2001).
29. Luo, C. & Rudy, Y. A Dynamic model of the cardiac
ventricular action potential- simulations of ionic
currents and concentration changes. Circ. Res. 74,
1071–1097 (1994).
30. Noble, D., Varghese, A., Kohl, P. & Noble, P. Improved
guinea-pig ventricular cell model incorporating a diadic
space, IKr and IKs, and length- and tension-dependent
processes. Can. J. Cardiol. 14, 123–134 (1998).
31. Noble, D. Modelling the heart: from genes to cells to
the whole organ. Science 295, 1678–1682 (2002).
32. Hunter, P. J. & Smaill, B. H. in Cardiac
Electrophysiology: from cell to bedside 3rd edn Vol. 32
(eds Zipes, D. P. & Jalife, J.) 277–283 (W. B. Saunders,
Philadelphia, 2000).
33. Tomlinson, K. A., Hunter, P. J. & Pullan, A. J. A finite element
method for an eikonal equation model of myocardial
excitation wavefront propagation. SIAM J. Appl. Math. 63,
324–350 (2002).
34. Hunter, P. J., McCulloch, A. D. & ter Keurs, H. E. D. J.
Modeling the mechanical properties of cardiac muscle.
Prog. Biophys. Mol. Biol. 69, 289–331 (1998).
35. Nash, M. P. & Hunter, P. J. Computational mechanics
of the heart. J. Elast. 61, 113–141 (2001).
36. Kohl, P., Hunter, P. J. & Noble, D. Stretch-induced
changes in heart rate and rhythm: clinical observations,
experiments and mathematical models. Prog. Biophys.
Mol. Biol. 71, 91–138 (1999).
37. Smith, N. P., Pullan A. J. & Hunter, P. J. Generation of
an anatomically based geometric coronary model.
Ann. Biomed. Eng. 28, 14–25 (2000).
38. Bradley, C. P., Pullan, A. J. & Hunter, P. J. Geometric
modeling of the human torso using cubic hermite
elements. Ann. Biomed. Eng. 25, 96–111 (1997).
39. Hunter, P. J., Robbins P. & Noble, D. The IUPS Human
Physiome Project. Pflugers Arch. Eur. J. Physiol. 445,
1–9 (2002).
40. Tawhai, M., Pullan, A. J. & Hunter, P. J. Generation of
an anatomically based three-dimensional model of the
conducting airways. Ann. Biomed. Eng. 28, 793–802
(2000).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 4 | MARCH 2003 | 243
Towards an e-biology of ageing:
integrating theory and data
Thomas B. L. Kirkwood, Richard J. Boys, Colin S. Gillespie,
Carole J. Proctor, Daryl P. Shanley and Darren J. Wilkinson
I NNOVAT ION
Ageing is a highly complex process; it involves
interactions between numerous biochemical
and cellular mechanisms that affect many
tissues in an organism. Although work on the
biology of ageing is now advancing quickly,
this inherent complexity means that
information remains highly fragmented. We
describe how a new web-based modelling
initiative is seeking to integrate data and
hypotheses from diverse biological sources.
Recent years have seen rapid progress in
understanding the science of ageing. A key
factor has been the interaction between evo-
lutionary (why?) and mechanistic (how?)
lines of research — this has helped to shape
the probable genetic basis of ageing and the
mechanisms that might be involved1. It has
also helped to overcome a situation in which
the field was dominated by a plethora of
rival theories with little effective dialogue
between them. In particular, the ‘disposable
soma theory’1,2 suggests that ageing is caused
by evolved limitations in organisms’ invest-
ments in somatic maintenance and repair,
rather than by active gene programming.
This predicts that ageing is due to the grad-
ual accumulation of unrepaired random
molecular faults, which leads to an increased
fraction of damaged cells and eventually to
© 2003 Nature Publishing Group

244 | MARCH 2003 | VOLUME 4 www.nature.com/reviews/molcellbio
P E R S P E C T I V E S
research for a concerted attempt to integrate
the findings from different lines of experi-
mental work.
Recognizing this need, about ten years ago
we began to build ‘network models’ of cellu-
lar ageing to examine, using computer simu-
lations, the candidate biochemical mechanisms
of ageing and their interactions. These models
have shown how several processes acting
together — often synergistically — can cause
cellular ageing, even though each mechanism
on its own might make only a modest contri-
bution6–10.We have shown that the predictions
from the network models are better able to
explain the experimental data than predictions
from individual single-cause theories.
Our models have also provided detailed,
quantifiable predictions of the changes in key
biochemical parameters of ageing cells over
time. These predictions provided a valuable
insight into the idea that, when the interactions
between different mechanisms were taken into
account, cellular instability and death could
result from relatively small changes in the bio-
chemical parameters of the cell. In other words,
the ‘upstream’ biochemical changes that are
ultimately responsible for age-related cellular
deterioration will often be hard to detect
experimentally. In addition, our models have
emphasized the importance of interactions
and stochastic events in explaining data, such
as the marked cell-to-cell variation in the divi-
sion potential of individual cells in the ‘replica-
tive-senescence’ model of cellular ageing, in
which normal cells are observed to undergo
only a finite number of divisions.
Despite the important contributions that
can be made by these kinds of modelling
studies, their usefulness is often limited
because experimenters are sceptical about the
value of mathematical models. Moreover,
unless the reader is an experienced modeller,
an important contribution to ageing that
arises from chance variations, which are not
explained by genetic or environmental fac-
tors3. A particularly clear example of the role
of chance in ageing is the threefold variation
in the lifespan, and the apparently stochastic
age-related cell degeneration, of individual
nematode worms that are in isogenic popu-
lations of Caenorhabditis elegans reared
under uniform laboratory conditions4,5.
Third, as the various mechanisms of ageing
probably interact, there should be a high
level of complexity in the overall mecha-
nisms of ageing. For all of these reasons,
there is an exceptional need in ageing
the functional impairment of older tissues
and organs. Genetic effects on the rate of
ageing are, according to this viewpoint,
mediated mainly through genes that influ-
ence somatic maintenance and repair.
Although the idea of ageing as a build-up
of damage is straightforward in principle
and supported by a growing range of data, it
presents several challenges. First, it predicts
that many mechanisms cause ageing, rather
than just one or a few. This is a powerful
argument because it predicts that each of the
many ‘single-cause’ theories is incomplete.
Second, it predicts that ageing is inherently
stochastic — there is extensive evidence for
Box 1 | Why model complex systems?
Modelling requires that verbal hypotheses be made specific and conceptually rigorous
• Before a mathematical model can be formulated, the investigator must specify each element of
the model and how it interacts with other elements.
Modelling highlights gaps in knowledge
• The process of specifying a mathematical model will highlight any important unknowns.
Sometimes these can be represented as variables yet to be estimated or determined.
• The recognition of a gap that needs to be filled by further experimental investigation might be
fundamental to understanding a complex system, even if the gap means that a model cannot yet
be completed.
Modelling provides quantifiable as well as qualitative predictions
• A hypothesis can be tested much more rigorously by a model that allows quantifiable
predictions to be made. For ageing, in which numerous mechanisms might be at work, data are
often broadly consistent with a hypothesized mechanism, but modelling can show that the
magnitude of the effect is too small to explain ageing on its own.
Modelling is ideal for analysing complex interactions before experimental tests
• Modelling might show that because of interactions within complex systems, a proposed
experiment would be inconclusive.
Modelling is a low-cost, rapid test bed for candidate interventions
• Modelling can allow a more predictive approach, effecting considerable savings in time and
money.
Well-designed models are readily portable and adaptable for many purposes
Modern theory of evolution
(why?) of ageing initiated by
Medawar42.
Williams extends evolutionary
theory by introducing the
concept of trade-offs44.
Free-radical theory of
ageing proposed by
Harman43.
Szilard begins to quantify the
idea that somatic mutations
might be a cause of ageing11.
Hayflick and Moorhead describe the
phenomenon of cell replicative senes-
cence, providing a major impetus to
work on the cellular basis of ageing45.
Disposable soma theory
bridges the evolutionary
(why?) and mechanistic
(how?) theories of ageing8.
Orgel suggests that cells might age
through a feedback of
random mistakes in macromolecule
synthesis12.
Network concept for mecha-
nisms of ageing described by
Kirkwood and Franceschi47.
First substantive network
theory of cellular ageing is
developed by Kowald and
Kirkwood7.
First single-gene mutation
extending lifespan is
described in a model
organism, Caenorhabditis
elegans46.
1952 1956 1957 1959 1961 1963 1977 1988 1992 1996
Timeline | Progress towards an integrated biology of ageing
© 2003 Nature Publishing Group

P E R S P E C T I V E S
(mt)DNA, occurring over years, might lead
to an insidious increase in the production of
reactive oxygen species and a gradual
decline in energy production7. However,
although the build-up of mtDNA mutations
initiates the process, the cell is destroyed
because a threshold is eventually reached
whereby homeostatic mechanisms collapse.
The end-stage of the cell’s lifespan is domi-
nated by marked biochemical changes, such
as an accumulation of damaged protein.
Phenomenological study of the latter effect
would not necessarily reveal the former
cause.
Another benefit of building an integrative
model is that it is well-suited to account for
the fact that many of the key reactions that are
involved in normal cell maintenance and
metabolism do not act in isolation — rather,
they belong to a network of activity. So, if the
activity of one enzyme changes then all of the
connected metabolite pools and enzyme
activities could be altered. In some cases, there
might be redundancy in the pathways that
buffer cells against damage, but in other cases
the effect of any damage might be propagated.
An effective theoretical and experimental
framework for analysing the control and reg-
ulation of metabolic systems is metabolic
control analysis (MCA)13,14. MCA is a formal
system of analysis that can examine how the
different components in a metabolic system
influence, and are influenced by, the concen-
trations and fluxes of metabolites through the
reaction pathways. MCA has been successfully
applied to understanding the control of mito-
chondrial bioenergy production15–17. A new
theory that has been developed for multilevel
it is hard to get a sufficient feel for the proper-
ties and behaviour of a model by reading a
published description.
For these reasons, we recently began a pro-
ject to create a web-based modelling system
that is known as BASIS (Biology of Ageing
e-Science Integration and Simulation). BASIS
aims to make both existing and new models
accessible to the research community in a way
that users can adapt and run themselves. The
aim is to develop BASIS to be a collective and
collaborative activity, which provides a frame-
work for other groups (both users and collab-
orators) to share in building a bioinformatics
resource that can help to integrate theory and
data on the biology of ageing. To this end,
BASIS will use the emerging technology of the
GRID — the significant enhancement of
global IT infrastructure that is now under way
— to enable enhanced sharing of distributed
computing and data resources.As BASIS seeks
to grow through widespread collaboration
and user involvement, much of the develop-
ment of BASIS has yet to be done, and this
article provides an overview of the work
ahead.
Why build models?
The value of mathematical and computer
models as predictive and analytical tools
for understanding complex biological
processes is increasingly recognized (BOX 1).
Nevertheless, many experimenters have yet
to be convinced of the usefulness of theo-
retical models, especially when they involve
many parameters. It is therefore important
to be explicit about the aims and objectives
of modelling a complex process like ageing.
These aims are: to improve the under-
standing of the biological process or
hypothesis that is under consideration; to
highlight gaps in the knowledge of the
process/hypothesis; and to be able to make
clear, testable predictions.
Large numbers of parameters are not
necessarily a problem if most of them can be
estimated reliably from data. Modelling
allows the study of what might be predicted if
a particular parameter — for example, an
enzyme level or reaction rate — is altered. It is
significant that, in the field of ageing, those
single-cause theories that have been most rig-
orously tested have tended to be those that
made the most specific quantifiable predic-
tions. So, the somatic-mutation theory, which
was modelled as early as 1959 (TIMELINE; REF.11),
was quickly criticized for not agreeing with
quantified data. Similarly, the protein-error
theory12, proposed in 1963 and modelled dur-
ing the 1970s, was subjected to numerous
quantifiable tests, which it generally failed.
Although these theories were largely dis-
missed as single-cause hypotheses, there is
evidence that both somatic mutations and
aberrant proteins accumulate during ageing
and might contribute in a causal way to cell
deterioration in normal ageing processes and
in age-related diseases, such as autoimmunity
and Alzheimer’s disease. Interestingly, theo-
ries that were less specific in their predictions
have tended not to be subjected to such rigor-
ous quantification.
Although the large number of ageing
mechanisms is now widely acknowledged,
the reductionist nature of experimental tech-
niques means that, in practice, most research
is still focused on single mechanisms. This is
where mathematical modelling can con-
tribute — by beginning the difficult, but
essential, task of putting the pieces together.
By allowing for interaction and synergism
between different processes, models show
that the predicted effects on the system are
often much greater than if mechanisms are
considered one at a time. Furthermore,
models can highlight important differences
between the upstream mechanisms that initi-
ate a process and the end-stage mechanisms
that dominate the cellular phenotype at the
end of its life (FIG. 1). For example, a gradual
accumulation of mutations in mitochondrial
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 4 | MARCH 2003 | 245
Direct evidence of func-
tional age changes
affecting tissue stem
cells was described in
intestinal epithelium34–36.
Bioinformatics tools are devel-
oped and applied through
BASIS to aid the sharing of
concepts and databases.
BASIS project initiated
to develop a GRID-
based integration and
simulation system for
the biology of ageing.
BASIS enables researchers
around the world to explore
interactions of different
candidate mechanisms of
ageing, generating a more
complete understanding of
this complex process.
1998 2002 Future
Figure 1 | The sequential development of ageing mechanisms. Modelling can reveal how the
mechanisms responsible for cellular ageing can develop sequentially as the process advances6. In
simulations, it is found that a long period of gradual accumulation of mitochondrial (mt)DNA mutations
results in a progressive increase in intracellular stress and a decline in energy production that eventually
precipitates a more pronounced breakdown of cellular homeostasis.
Mutation-free
mitochondria
Accumulation
of mutant
mitochondria
End-stage of
cellular
collapse
Gradual accumulation of
mtDNA mutations
Decreased ATP production
and increased oxidative
stress precipitate
homeostatic failure
Slow Fast
“…there is an exceptional
need in ageing research for
a concerted attempt to
integrate the findings from
different lines of
experimental work.”
© 2003 Nature Publishing Group

246 | MARCH 2003 | VOLUME 4 www.nature.com/reviews/molcellbio
P E R S P E C T I V E S
The web service will be implemented so
that users can access BASIS facilities using a
conventional web browser and, for model
development, the system will download
‘client software’ that allows users to modify
models by using a flexible graphical user
interface (FIG. 2). The user interface, which is
under development, will allow users to build
up or alter a model using a series of ‘click-
and-drag’ operations (similar to the way
that a diagram might be drawn), and each
element will have a set of associated para-
meter values that can be accessed and
changed at will. The model will be translated
into appropriate code using the systems biol-
ogy markup language (SBML) — a descrip-
tion-language for simulations, especially of
biochemical networks (see online links). The
translation process and its output will nor-
mally be invisible to the user, unless specified
otherwise, and results will be plotted in
appropriate windows (for example, as
graphs of biochemical variables changing
over time) or sent as files. In addition to the
modelling software, BASIS will also support
databases of relevant parameter values, liter-
ature and other models. Within the overall
web service, BASIS will provide utilities to
support the exchange of ideas and dialogue
among modellers and experimenters.
The main modelling modules of BASIS
will provide representations of systems at
three levels: cell, tissue and organism. These
will serve two main user groups: biologists
who wish to explore and use the available
models; and modellers, who might be inter-
ested in the role of specific candidate mecha-
nisms that are not yet included in the BASIS
model-set, and who wish to explore these
mechanisms in a framework that will allow
any potential interactions with the processes
that are already in the BASIS model-set to be
taken into account. It will be through these
kinds of collaborations that additions to the
BASIS model-set will be made.
Virtual ageing cell
So far, most of the modelling work has focused
on intracellular mechanisms that result in the
degeneration and death of an ageing cell. We
have shown how different mechanisms interact
synergistically (for example, the interactions of
mtDNA mutations, free-radical production
and the intracellular accumulation of aberrant
proteins).We have also shown how the relative
importance of individual components of the
network is influenced by factors such as
whether the cell continues to divide on a reg-
ular basis (for example, a fibroblast) or not
(for example, a neuron). Indeed, our model
of mitochondrial dynamics26 predicts very
can be expected to respond to perturbation
by random molecular damage that affects one
or more of its modules17.
MCA is also important when considering
possible interventions in ageing processes,
such as attempting to slow or even arrest the
damage that is caused by oxidative stress. For
example, increasing the concentration of a
single enzyme in the chain of cellular antioxi-
dant defences, such as superoxide dismutase
(SOD), could do more harm than good if it
perturbs the balance of intermediate metabo-
lites. For example, SOD captures the damag-
ing superoxide radicals and, as a reaction
product, generates hydrogen peroxide, which
is also an oxidizing radical that must be ren-
dered harmless by a second enzyme called
catalase. Experiments in fruit flies that used
transgenes to boost SOD activity showed that
a positive effect on lifespan was seen only
when the amount of catalase was increased at
the same time as the amount of SOD19.
Another important area for modelling is
to understand the actions of genes that
affect the rate of ageing. Over the past
decade, many genes have been identified
that affect ageing in yeast, nematodes, fruit
flies and mice, and there is growing interest
in genes that affect human longevity20–23.
Experimental data are beginning to reveal the
interactions of these genes in pathways that
control the ageing rate. There is evidence that
several of the most important genes are those
that affect basic cellular processes, such as
insulin and insulin-like growth factor (IGF)
signalling, which are strongly conserved
across species24. Nevertheless, we are a long
way short of understanding the interactions
between these effects. The relevant genes
merely modulate (but do not abolish) the
processes that cause cells and organisms to
age, and they probably act through factors
that influence the rate of accumulation of
somatic-cell damage. These studies need also
to take account of the intrinsic stochastic
nature of gene regulatory networks25.
Aims and scope of the BASIS project
BASIS has two main aims: to significantly
extend the scope of current integrative mod-
els and to make these models widely accessi-
ble through the internet. Accessibility will be
achieved by developing a web service, through
which investigators can explore models and
run simulations for themselves, as well as
reviewing results that have been run by others
and placed in the public areas of the system.
This will help to overcome a serious restric-
tion of the conventional publication of pre-
dictions from models, which is limited to just
a few illustrative examples.
reaction networks18 has made it possible to
dissect cascades of control processes quantifi-
ably, addressing questions such as how the
control properties of the reaction network as
a whole can be expressed in terms of the con-
trol properties of individual reactions and
interaction between reactions. So far, work
has concentrated on understanding the regu-
latory properties of the normal cell. However,
it is immediately clear that the framework of
MCA and multilevel reaction networks will
be important in examining how the system
Figure 2 | Organizational scheme for the
BASIS project. a | The core GRID node will be in
Newcastle. It will be supported by a dedicated
high-performance-computer cluster with further
software support from the North East Regional
e-Science Centre. Collaborators will share in the
development of models and databases that will
be integrated with BASIS through the GRID (the
first international BASIS collaborators’ workshop
took place in November 2002). b | Users can
access the web service through a conventional
web browser. Alternatively, BASIS client software
will run on the user’s PC for model specification.
For simulations that are too computationally
demanding to run interactively, users will receive
automatic e-mail notification when their results are
available. API, application programmer interface.
Collaborators
Modelling, databases
Core GRID node
Modelling,
databases,
compute server
Users
Data, ideas
Job
schedulerDatabase
API
Web services
CGI scripts
Web server
Web browserUser PC BASIS client
software
e-mail
notification
BASIS
file
server
Internet
Linux beowulf cluster
a
b
© 2003 Nature Publishing Group

P E R S P E C T I V E S
The model will be an isotropic three-
dimensional matrix of cells, each of which
will be subject to stochastic ageing processes.
When they are surrounded by other cells,
fibroblasts divide only rarely, but occasional
cell deaths in the matrix will create gaps that
need to be filled by the division of neighbour-
ing cells. For example, an individual cell
might, by chance, accumulate a high level of
defective mitochondria — those containing
mutations in the mtDNA that encodes key
respiratory enzymes — which, in turn, might
trigger apoptosis (programmed cell death).
The connective-tissue model will also allow us
to investigate how senescent cells in a tissue can
have significant effects on tissue properties, for
example, by secreting increased levels of
enzymes that degrade the extracellular collagen
matrix32.
The next two models will consider tissues
that are more and less proliferative than con-
nective tissue. One will be based on intestinal
epithelium — the most proliferative tissue in
the mammalian body — for which there are
extensive data on the patterns and rates of
cell division in the normal, young animal. In
particular, the functional homeostasis of this
different outcomes for the accumulation of
mtDNA mutations in dividing and non-divid-
ing cells, and this has been observed experi-
mentally27,28.
BASIS models will include important
interactions between individual mecha-
nisms that are observed experimentally. For
example, it is well known that telomeres in
telomerase-negative somatic cells shorten
at each cell division, and there is evidence
that telomere shortening is associated with
replicative senescence. In particular, it has
been shown that the telomere-shortening
rate is accelerated by oxidative stress29.
Another important observation is the intra-
mitotic and intra-clonal variation in the
population-doubling potential of human
diploid fibroblasts30,31; this stochastic varia-
tion will also be included in the BASIS
models. The capacity of models to explain
stochastic variation both qualitatively and
quantifiably is important (FIG. 3).
The ‘virtual ageing cell’ will allow key
processes to be represented in relatively sim-
ple terms or to be expanded into more
detailed structures, as hypotheses or knowl-
edge allow. One example is the representa-
tion of the cell’s antioxidant enzyme
defences. For some purposes it is enough to
include simply a generic antioxidant enzyme
that represents the overall activity of the
antioxidant complex. For other purposes —
for example, where the effect of upregulating
a particular enzyme is of interest — greater
detail is needed. As a mathematical model is
always a simplification of reality, a judge-
ment must be made about which biological
and mathematical features to include and
which to leave out. Such judgements could
leave the non-mathematical biologist unsure
about whether or not the model can be
trusted. BASIS will expose more of this
process to the biological user, allowing the
user to learn to make judgements that are
based on greater insight and understanding,
and so modellers can be kept informed by
feedback from experimenters.
The virtual ageing cell will simulate the
behaviour of individual cells according to the
processes (stochastic or deterministic) that
are incorporated into the cellular structure
(BOX 2). It will also enable the cell’s prolifera-
tive behaviour to be specified, allowing not
only for the presence/absence of cell division,
but also for the possibility of asymmetric cell
division, which occurs, for example, in bud-
ding yeast (an important model system for
the study of ageing).
Virtual ageing tissues
The functional properties of an ageing organ
or tissue can become compromised, even if
most of the cells are in good working order.
For example, marked changes are seen in aged
human connective tissue, but cells that are cul-
tured from skin biopsies of human centenari-
ans grow with almost the same vigour as those
cultured from young adults because most of
the cells are still in good condition. To try and
understand how a fraction of damaged cells
can lead to altered tissue function, the second
level of BASIS will model virtual aged tissues.
The simplest model will comprise a
matrix of connective-tissue cells (fibroblasts).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 4 | MARCH 2003 | 247
Figure 3 | Stochastic modelling of cell replicative senescence. The figure shows a comparison
between a–c | experimental data and d–f | simulation. The data represent heterogeneity in the
cell-doubling potential of individual fibroblasts that were sampled at population-doubling levels of 39 (a),
49 (b) and 59 (c) from a mass population that had been established by clonal expansion from a single
founder cell30. The simulations (d–f) are from a model that is based on interactions between telomere
shortening, oxidative stress and somatic mutations in nuclear and mitochondrial DNA, which mimics the
experimental procedures9.
Fr
eq
ue
nc
y
50
40
30
20
10
0
0 5 10 15 20 25 30
50
40
30
20
10
0
0 5 10 15 20 25 30
50
40
30
20
10
0
0 5 10 15 20 25 30
50
40
30
20
10
0
0 5 10 2015 25 30 35 40
50
40
30
20
10
0
0 5 10 2015 25 30 35 40
50
40
30
20
10
0
0 5 10 2015 25 30 35 40
Number of cell population doublings
a d
b e
c f
“The particular union of
disciplines that underpin
the BASIS project — ageing
and mathematical
modelling — has been
singled out for recognition
as an exemplar of
multidisciplinary science.”
© 2003 Nature Publishing Group

248 | MARCH 2003 | VOLUME 4 www.nature.com/reviews/molcellbio
P E R S P E C T I V E S
underpin the BASIS project — ageing and
mathematical modelling — has been singled
out for recognition as an exemplar of multi-
disciplinary science41.
Looking forward — towards the applica-
tion of functional genomics and other fast-
moving technologies to major challenges in
biology, including the growth in prevalence of
age-related degenerative conditions — it is
clear that we are embarking on a trend
towards a more predictive kind of science.
The BASIS project is intended to contribute
to this process. Information about the
progress of the project, and how to participate
in and use it, will be available through the
BASIS web site.
Thomas B. L. Kirkwood, Colin S. Gillespie, Carole J.
Proctor and Daryl P. Shanley are at the Institute
for Ageing and Heath, University of Newcastle,
Newcastle General Hospital, Newcastle upon Tyne
NE4 6BE, UK.
Richard J. Boys and Darren J. Wilkinson are at the
School of Mathematics and Statistics, University of
Newcastle, Newcastle upon Tyne NE1 7RU, UK.
Correspondence to T.B.L.K.
e-mail: Tom.Kirkwood@ncl.ac.uk.
doi:10.1038/nrm1051
1. Kirkwood, T. B. L. & Austad, S. N. Why do we age?
Nature 408, 233–238 (2000).
2. Kirkwood, T. B. L. Evolution of ageing. Nature 270,
301–304 (1977).
3. Finch, C. E. & Kirkwood, T. B. L. Chance, Development
and Aging (Oxford Univ. Press, New York, 2000).
4. Herndon, L. A. et al. Stochastic and genetic factors
influence tissue-specific decline in ageing C. elegans.
Nature 419, 808–814 (2002).
5. Kirkwood, T. B. L. & Finch, C. E. The old worm turns
more slowly. Nature 419, 794–795 (2002).
6. Kowald, A. & Kirkwood, T. B. L. Mitochondrial mutations,
cellular instability and ageing: modelling the population
dynamics of mitochondria. Mutat. Res. 295, 93–103
(1993).
7. Kowald, A. & Kirkwood, T. B. L. A network theory of
ageing: the interactions of defective mitochondria,
aberrant proteins, free radicals and scavengers in the
ageing process. Mutat. Res. 316, 209–236 (1996).
8. Kirkwood, T. B. L & Kowald, A. Network theory of ageing.
Exp. Gerontol. 32, 395–399 (1997).
9. Sozou, P. D. & Kirkwood, T. B. L. A stochastic network
model of cell replicative senescence based on telomere
shortening, oxidative stress and somatic mutations in
nuclear and mitochondrial DNA. J. Theor. Biol. 213,
573–586 (2001).
10. Proctor, C. & Kirkwood, T. B. L. Modelling telomere
shortening and the role of oxidative stress. Mech. Ageing
Dev. 123, 351–363 (2002).
11. Szilard, L. On the nature of the aging process. Proc. Natl
Acad. Sci. USA 45, 35–45 (1959).
12. Orgel, L. E. The maintenance of the accuracy of protein
synthesis and its relevance to ageing. Proc. Natl Acad.
Sci. USA 49, 517–521 (1963).
13. Kacser, H. & Burns, J. A. The control of flux. Symp. Soc.
Exp. Biol. 7, 65–104 (1973).
14. Kacser, H. & Burns, J. A. Molecular democracy: who
shares the controls? Biochem. Soc. Trans. 7, 1149–1160
(1979).
15. Letellier, T, et al. Metabolic control analysis and
mitochondrial pathologies. Mol. Cell. Biochem. 184,
409–417 (1998).
16. Ainscow, E. K. & Brand, M. D. Top-down control analysis
of ATP turnover, glycolysis and oxidative phosphorylation
in rat hepatocytes. Eur. J. Biochem. 263, 671–685
(1999).
17. Murphy, M. P. How understanding the control of energy
metabolism can help investigation of mitochondrial
dysfunction, regulation and pharmacology. Biochem.
Biophys. Acta. 1504, 1–11 (2001).
successfully used model organisms, the
nematode C. elegans. This species has a pre-
cisely determined body plan, comprising just
959 somatic cells in the adult, which is a com-
putationally manageable number. Each cell
can be identified in the developmental lineage,
and functional genomics studies in C. elegans
are now well advanced, including studies of
targeted cell disruption37,38. Although these
features make C. elegans an attractive and
tractable model system to begin the process
of modelling a virtual ageing organism, it is
important to bear in mind that aged worms
are not tiny aged people. One limitation of
C. elegans as a model for ageing in mammals
is the absence of cell division in the adult, so
the important role of cell proliferation and
its perturbation with age will not yet be
represented in this virtual ageing organism.
Towards a predictive biology of ageing
The BASIS project exemplifies the radical
changes that have occurred in biology recently,
and they look set to accelerate39. When the
background stages of this work were begun,not
only was ageing research itself unfashionable,
but little mathematical modelling of cellular
and molecular processes was done. Scepticism
about ageing as a subject for serious scientific
research was put down to the unfathomable
complexity of the process40, whereas attitudes
to modelling were influenced both by unfa-
miliarity with the methodology and by a
common misconception that it was either
purely descriptive or wildly hypothetical.
Recently, however, enormous advances have
been made in our understanding of ageing.
The particular union of disciplines that
tissue is derived from small numbers of tissue
stem cells, the properties of which hold the
key to long-term maintenance and which
have been explored using stochastic models
of stem-cell organization33. Recently, we and
our collaborators generated extensive experi-
mental data on intrinsic age changes that
affect the function of intestinal stem cells in
ageing mice. These changes include:
increased sensitivity to DNA-damaging
agents; impaired ability to regenerate tissue
damaged by irradiation; and altered intracel-
lular mechanisms for responding to and
repairing DNA damage34–36. The model of
gut epithelium will provide an opportunity
to examine the effects of intrinsic cell ageing
on a proliferative tissue with a clearly defined
cell hierarchy, and in which cell migration
plays a key role.
The third model will represent the neuronal
networks that make up brain tissue. It will sim-
ulate the effects of ageing on a non-dividing tis-
sue in which function depends on the integrity
of intercellular connections. Redundancy in
neural networks means that learning and
memory functions are not necessarily destroyed
when a single cell is lost, and the model is
intended to help understand the impact of
intrinsic cellular ageing on the integrity of learn-
ing and memory.
Virtual ageing organism
The most demanding phase of BASIS will
aim to integrate the elements of the ageing
process at the level of a whole organism.
Achieving this goal for an organism as com-
plex as a mammal is unrealistic at present.
However, it is feasible for one of the most
Box 2 | Building the virtual ageing cell
The virtual ageing cell is both a model and a tool. It aims to encourage newcomers to modelling
to modify existing models and run them with user-selected parameter values. Few of the
intended users are likely to have mathematical programming skills. They will, however, be used
to using the term ‘model’ to describe a diagram that represents how they think a biological
system works. The user interface is being constructed so that models can be built up or edited
in the same way as these kinds of diagrams. The difference is that the user will need to put
numbers to the elements and their interactions in the diagram. The user will also specify what
kind of cell is to be studied (for example, dividing versus post-mitotic) and which endpoints
and outputs are of interest (for instance, time to cell death or the evolution of key biochemical
parameters).
The virtual ageing cell is being developed using a set of interacting modules to represent key
variables and reaction pathways within the cell. A well-supported approach based on XML and
web services technology is being developed within the broader framework of the UK e-science
initiative.
A combination of deterministic and stochastic models will be included. Deterministic models
are appropriate, for example, when using simultaneous differential equation simulation of
enzyme reactions in which numerous reactant molecules are involved. Stochastic models are
needed, for example, for simulation of the probabilistic processes that underlie mutation and the
intrinsic variability in the control of gene expression. To these will be added a representation of
state-dependent reaction fluxes between metabolic compartments.
© 2003 Nature Publishing Group

P E R S P E C T I V E S
40. Wolpert, L. The Unnatural Nature of Science 135–136
(Faber and Faber, London, 1992).
41. Editorial. Nature 417, 471 (2002).
42. Medawar, P. B. An Unsolved Problem of Biology (H. K.
Lewis, London, 1952).
43. Harman, D. Aging: a theory based on free radical and
radiation chemistry. J. Gerontol. 11, 298–300 (1956).
44. Williams, G. C. Pleiotropy, natural selection and the
evolution of senescence. Evolution 11, 398–411 (1957).
45. Hayflick, L. & Moorhead, P. S. The serial cultivation of human
diploid cell strains. Exp. Cell Res. 25, 585–621 (1961).
46. Friedman, D. B. & Johnson, T. E. Three mutants that
extend both mean and maximum life-span of the
nematode, Caenorhabditis elegans, define the age-1
gene. J. Gerontol. 43, B102–B109 (1988).
47. Kirkwood, T. B. L. & Franceschi, C. Is aging as complex as
it would appear? Ann. NY Acad. Sci. 663, 412–417 (1992).
Acknowledgements
The BASIS project is funded by the UK Biotechnology and
Biological Sciences Research Council, Medical Research Council
and Department of Trade and Industry. Our work is also funded by
the BBSRC Science of Ageing initiative.
Online links
FURTHER INFORMATION
Thomas B. L. Kirkwood’s laboratory:
http://www.ncl.ac.uk/medi/research/gerontology
BASIS project: http://www.basis.ncl.ac.uk
Systems biology markup language:
http://www.sbw-sbml.org
e-science initiative: http://www.nesc.ac.uk
Access to this interactive links box is free online.
18. Hofmeyr, J. S. & Westerhoff, H. V. Building the cellular
puzzle: control in multi-level reaction networks. J. Theor.
Biol. 208, 261–285 (2001).
19. Orr, W. C. & Sohal, R. Extension of life span by
overexpression of superoxide dismutase and catalase in
Drosophila melanogaster. Science 263, 1128–1130
(1994).
20. Jazwinski, S. M. Metabolic mechanisms of yeast ageing.
Exp. Gerontol. 35, 671–676 (2000).
21. Lithgow, G. J. Aging mechanisms from nematodes to
mammals. Nutrition 14, 522–524 (1998).
22. Larsen, P. L. Asking the age-old questions. Nature
Genet. 28, 102–104 (2001).
23. Gems, D. & Partridge, L. Insulin/IGF signalling and
ageing: seeing the bigger picture. Curr. Opin. Genet. Dev.
11, 287–292 (2001).
24. Cournil, A. & Kirkwood, T. B. L. If you would live long,
choose your parents well. Trends Genet. 17, 233–235
(2001).
25. McAdams, H. H. & Arkin, A. Stochastic mechanisms in
gene expression. Proc. Natl Acad. Sci. USA 94, 814–819
(1997).
26. Kowald, A. & Kirkwood, T. B. L. Accumulation of
defective mitochondria through delayed degradation of
damaged organelles and its possible role in the ageing of
post-mitotic and dividing cells. J. Theor. Biol. 202,
145–160 (2000).
27. Cortopassi, G. A., Shibata, D., Soong N. W. & Arnheim, N.
A pattern of accumulation of a somatic deletion of
mitochondrial-DNA in aging human tissues. Proc. Natl
Acad. Sci. USA 89, 7370–7374 (1992).
28. Lee, C. M., Pang, C. Y., Hsu, H. S. & Wei, Y. H.
Differential accumulation of 4977 bp deletion in
mitochondrial DNA of various tissues in human ageing.
Biochim. Biophys. Acta. 1226, 37–43 (1994).
29. von Zglinicki, T., Bürkle, A. & Kirkwood, T. B. L. Stress,
DNA damage and ageing — an integrative approach.
Exp. Gerontol. 36, 1049–1062 (2001).
30. Holliday, R., Huschtscha, L. I., Tarrant, G. M. & Kirkwood,
T. B. L. Testing the commitment theory of cellular ageing.
Science 198, 366–372 (1977).
31. Smith, J. R. & Whitney, R. G. Intraclonal variation in
proliferative potential of human diploid fibroblast cells:
stochastic mechanism for cellular aging. Science 207,
82–84 (1980).
32. Campisi, J. From cells to organisms: can we learn about
aging from cells in culture? Exp. Gerontol. 36, 607–618
(2001).
33. Loeffler, M. et al. Somatic mutation, monoclonality and
stochastic models of stem cell organization in the
intestinal crypt. J. Theor. Biol. 160, 471–491 (1993).
34. Martin, K., Kirkwood, T. B. L. & Potten, C. S. Age
changes in stem cells of murine small intestinal crypts.
Exp. Cell Res. 241, 316–323 (1998).
35. Martin, K., Potten, C. S., Roberts, S. A. & Kirkwood, T. B. L.
Altered stem cell regeneration in irradiated intestinal
crypts of senescent mice. J. Cell Sci. 111, 2297–2303
(1998).
36. Martin, K., Potten, C. S. & Kirkwood, T. B. L. Age-related
changes in irradiation-induced apoptosis and expression
of p21 and p53 in crypt stem cells of murine intestine.
Ann. NY Acad. Sci. 908, 315–318 (2000).
37. Harbinder, S. et al. Genetically targeted cell disruption in
Caenorhabditis elegans. Proc. Natl Acad. Sci. USA 94,
13128–13133 (1997).
38. Arantes-Oliveira, N., Apfeld, J., Dillin, A. & Kenyon, C.
Regulation of life-span by germ-line stem cells in
Caenorhabditis elegans. Science 295, 502–505 (2002).
39. Alliance for Cellular Signaling. Overview of the Alliance for
Cellular Signaling. Nature 420, 703–706 (2002).
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 4 | MARCH 2003 | 249
© 2003 Nature Publishing Group