Systems Biology and Its Application to the
Understanding of Neurological Diseases
Pablo Villoslada, MD,1
Lawrence Steinman, MD,2
and Sergio E. Baranzini, PhD3
Recent advances in molecular biology, neurobiology, genetics, and imaging have demonstrated important insights about the
nature of neurological diseases. However, a comprehensive understanding of their pathogenesis is still lacking. Although reductionism
has been successful in enumerating and characterizing the components of most living organisms, it has failed to generate
knowledge on how these components interact in complex arrangements to allow and sustain two of the most fundamental
properties of the organism as a whole: its fitness, also termed its robustness, and its capacity to evolve. Systems biology complements
the classic reductionist approaches in the biomedical sciences by enabling integration of available molecular, physiological,
and clinical information in the context of a quantitative framework typically used by engineers. Systems biology employs
tools developed in physics and mathematics such as nonlinear dynamics, control theory, and modeling of dynamic systems. The
main goal of a systems approach to biology is to solve questions related to the complexity of living systems such as the brain,
which cannot be reconciled solely with the currently available tools of molecular biology and genomics. As an example of the
utility of this systems biological approach, network-based analyses of genes involved in hereditary ataxias have demonstrated a set
of pathways related to RNA splicing, a novel pathogenic mechanism for these diseases. Network-based analysis is also challenging
the current nosology of neurological diseases. This new knowledge will contribute to the development of patient-specific therapeutic
approaches, bringing the paradigm of personalized medicine one step closer to reality.
Ann Neurol 2009;65:124–139
“Information is not knowledge.”—Albert Einstein
Systems Biology: Underlying Principles,
Approaches, and Applications
Although a formal definition of systems biology has
not yet been widely accepted, most researchers agree in
that it represents an integrative approach that attempts
to understand higher-level operating principles of living
organisms, including humans.1,2
In biomedical research,
not all scientific questions are directly accessible
to experimentation. There is a hierarchy of scientific
questions whose levels are determined by the generality
of the answers sought. The lower levels in this hierarchy
deal with narrowly restricted and specific phenomena.
These are the kinds of questions that the reductionist
experimenter feels more comfortable in
addressing. On the other end of the spectrum, higherorder,
abstract, and general questions are not usually
directly amenable to an experimental test. Reductionist
thinking mandates that these questions be broken
down into more specific terms that can be translated
directly from experimental results.3
In stark contrast,
systems biology strives to understand these higherorder
properties whereas examining the complexity of
the system under study. Systems biology heavily relies
on the construction, utilization, and integration of inductive
models, a creative process for which a considerable
degree of abstraction is crucial. In this sense, abstraction
consists of replacing the part of the system
under consideration by a model of similar but simpler
structure.
Traditionally, neuroscientists have coped with the
enormous complexity of the brain in a purely reductionist
way: by subdividing it into anatomic regions
and characterizing each of their cellular compositions
and basic functions in isolation. Although that approach
has enjoyed remarkable success, the next challenge
is to translate such valuable information into a
better understanding of how several higher-order properties
of the brain, such as memory, learning, and behavior,
emerge from such a complex interplay (see the
Table for a glossary of terms). Under the reductionist
paradigm, a positive correlation between a single biological
parameter and the occurrence of a disease is ofFrom
the 1
Department of Neuroscience, Hospital Clinic–Institut
d’Investigacions Biome`diques August Pi i Sunyer, Barcelona, Spain;
2
Department of Neurology and Neurological Sciences, Stanford
University, Stanford; and 3
Department of Neurology, University of
California, San Francisco, CA.
Address correspondence to Dr Villoslada, Department of Neurology,
Hospital Clinic – IDIBAPS, Villaroel 170, 08036 Barcelona,
Spain. E-mail: pvilloslada@clinic.ub.es
Potential conflict of interest: Nothing to report.
Received Dec 21, 2007, and in revised form Dec 16, 2008. Accepted
for publication Dec 16, 2008.
Published online in Wiley InterScience (www.interscience.wiley.com).
DOI: 10.1002/ana.21634
NEUROLOGICAL PROGRESS
124 © 2009 American Neurological Association
Published by Wiley-Liss, Inc., through Wiley Subscription Services
ten considered a major success, even though the complete
pathogenic mechanism may remain largely
unknown. Notably, this approach has been followed by
the pharmaceutical industry, despite a remarkably low
rate of success (only 11% of new therapeutic targets
reach the market as new drugs).4
Systems biology focuses
on understanding not only the components of a
given system, thus complementing the reductionist approach,
but also the effect of interactions among them
and the interaction of the system with its environment.
Like physiology, systems biology is deeply rooted in
the principle that the whole is more than the sum of its
parts.5
Biological systems offer multiple examples of collective
properties, those in which the behavior of the
whole cannot be predicted from the detailed study of
individual components. In systems biology, as in the
science of complexity, those intrinsic properties are referred
to as emergent.6
Emergent properties are present
in all physiological systems and include the maintenance
of blood volume, blood pressure, tissue pH, or
body temperature. Emergent properties in the brain include
the processes of learning, memory, and emotions
that cannot be explained fully even by the detailed
study of single neurons. Analyses based on single-cell
neurobiology or molecular biology are useful for characterizing
the individual behavior of each component.
However, these properties are the result of nonrandom
interactions of highly specialized cells assembled in
large networks, and they can be understood only by
not disrupting these structures. Based on this and other
examples of systems composed of interconnected elements
(eg, acquaintances among people, the Internet,
airports), inductive models have been created that display
surprisingly generalizable properties.7,8
Although systems biology strives to combine techniques
obtained from seemingly disparate disciplines, it
mainly represents a way of thinking.9
Traditionally, the
study of body dynamics, as functional systems, has
been the focus of physiology. In this sense, systems biology
can be regarded as contemporary physiology, one
that uses methods from mathematics, physics, and
computer sciences to integrate and analyze genetic,
molecular, physicochemical, functional, and behavioral
data obtained in the laboratory.10,11
A systems approach
to research already has been established in several
fields of biology, such as ecology, enzymology, and
computational neuroscience. Such an approach is necessary
by virtue of the large amounts of data generated
by high-throughput methods such as gene chips and
proteomics. Moreover, we now realize that the human
genome is more complex and has many more levels of
regulation than previously anticipated,12,13
and systems
biology is expected to provide a comprehensive interpretation
of its organization.
Robustness and Disease
Two of the most critical emergent properties of all living
systems are robustness and adaptability.8,14
Robustness
refers to the ability of a system to maintain its
basic functions even in the presence of perturbations,
such gene mutations or environmental fluctua-
tions.15–17
However, when robustness alone is unable
to cope with such challenges, organisms must be able
to adapt to them and to evolve. Understanding these
higher-order properties requires the formulation of a
series of inductive analogues or models that capture the
critical aspects of the system and neglect the details,
thus shedding some of the initial complexity and making
it amenable to study.
One example of such an approach is systems control
theory, developed and used for decades by engineers
and physicists in the construction of advanced technological
equipment, ranging from thermostats to the radar
and guidance systems on airplanes.9,18
Systems
control theory has been successfully adapted to the detailed
study of the delicate balance between robustness
and adaptability in living systems.19
In thinking like
engineers, biologists have been able to create machinelike
analogues that faithfully explain the properties under
study. Through the use of these metaphors, such
studies have shown that robustness in life relies on several
fundamental principles. First, fail-safe mechanisms
such as redundancy and diversity (eg, gene duplication
or overlapping pathways) enable the organism to function
even if one of its molecules or pathways is affected
to a reasonable degree. Second, positive and negative
feedback and feed-forward mechanisms result in properties
such as multistability, oscillations, and signal amplification,
respectively.16,17
These mechanisms bestow
an organism with control functions such as ON/OFF
switches and autoregulation (homeostasis). Third, the
networked architecture of internal components allow
task parallelization, thus containing local damage and
preventing its spreading to the full system. Finally,
functional and physical properties, although related,
work independently (decoupling). In this way, changes
at the physical level (ie, protein misfolding) may not
translate to the functional level, resulting in pathology.
Chaperones are designed to deal with misfolding and
with unstable intermediate states in the final protein
structure.
By integrating these pieces of our inductive model
into a “general solution,” diseases can be viewed as a
breakdown in the robustness of normal physiological
systems. Chronic diseases would represent the establishment
of a new, but undesired, robust state.20
Advantages of a Systems Biology Approach to
Neurological Diseases
Computational neuroscience, neuroinformatics, neurophysiology,
and other newly developed disciplines are
Villoslada et al: Systems Biology in Neurology 125
Table. Concepts and Examples in Systems Biology
Concept Description and Example
System2,9
A set of connected entities forming an integrated whole. e.g., The cell is a system composed by
molecules (entities) with well-defined properties (division, metabolism, among others); the body is
another system composed by cells
Systems thinking9
Systems thinking is a framework that is based on the belief that the component parts of a system
can best be understood in the context of relations with each other and with other systems, rather
than in isolation. e.g., Explaining the physiology using feedback exemplified the attempt to
understand the influences of some biological process with each other
Complexity
theory2,9,57
A discipline of physics studying complex systems by focusing on the generation of emergent
properties in systems; complexity is not equivalent to complicated. e.g., Self-organization refers to
the generation of complex structures with well-defined properties without a master program (ie,
generation of liposomes when lipids are mixed with water) (see also Emergent properties and
Robustness)
Nonlinear
dynamics10
A discipline of physics that studies how systems evolve with time, involving a lack of linearity; a
nonlinear system is any problem where the variable(s) to be solved for cannot be written as a
linear sum of independent components. e.g., The majority of systems in nature are nonlinear
(linear analysis is a simplification); electroencephalographic signals are typically nonlinear because
they are the result of complex interactions between neurons and not just the sum of each
neuron’s activity
Control theory12
A discipline of engineering that deals with the behavior of dynamical systems. e.g., The analysis
of molecular networks as a set of positive and negative feedbacks, mimicking the analysis of an
electric circuit
Feedback
control1,2,19,43
A process whereby some proportion of the output signal of a system is passed (fed back) to the
input. e.g., Cortisol regulation by the hypothalamus-pituitary-adrenal axis
Robustness15–17
Maintenance of the properties and function of a system in the presence of perturbations. e.g.,
Homeostasis is a classic physiological example: the body maintains pH, glucose, temperature, and
so forth in the presence of changes in the environment (feeding, weather, among others)
Adaptability
14,15
A positive characteristic of an organism that has been favored by natural selection; sometimes also
called fragility in comparison with robustness. e.g., Genetic mutations are rare but present in
human genome allowing adaptation to the environment (ie, thalassemia mutation protects against
malaria)
Emergent
properties2,10
Physical properties of a system arising as an effect of complex causes and not analyzable simply
as the sum of their effects. e.g., Glucose levels stability is the result of many interacting
subsystems in the body (food intake, insulin, glucagons, cortisol, and so on); cognition is an
emergent property of the brain
Network7,8
Any interconnected group or system that shares information (ie, Internet, social networks). e.g.,
Gene networks are the set of genes that, through their proteins, interact with each other;
neuronal networks are the set of neurons connected through synapses participating in common
tasks (ie, learning)
Network analysis8
I. Network architecture (topology):
1. Microscale. e.g., Motif: recurring circuits of interactions between nodes (ie, gene interaction
loop)
2. Mesoscale. e.g., Module: group of physically or functionally linked molecules (nodes) that
work together to achieve a distinct function (ie, cell-cycle genes)
3. Macroscale. e.g., Scale free: only a few nodes are highly connected (hubs), and the majority
of the nodes are poorly connected (eg, Internet)
II. Network dynamics:
1. Mathematical modeling with differential equations
2. Synchronization analysis
3. Feedback analysis
126 Annals of Neurology Vol 65 No 2 February 2009
devoted to the study of emergent properties of the
brain.21,22
Although approaches aimed at deciphering
how the brain operates are gradually migrating from
deterministic to systems based, the two approaches currently
coexist somewhat separately.23
Although efforts
for identifying gene variants associated with intelligence,
memory, or social skills are currently under way,
neuroscientists acknowledge that the informationprocessing
capabilities and ultimate brain behavior result
from the dynamic interplay of complex networks
of synapses.
The identification of a gene defect associated with a
given disease is often followed by attempts to develop
therapies aimed at correcting its function, with the
hope that this may restore homeostasis. For example,
our current understanding of familial Alzheimer’s disTable.
(Continued)
Concept Description and Example
Network topology8
The architecture of nodes and links in a network that defines how information flows through the
network:
Nodes: elements of a network. Node: a gene, protein, or cell
Links: interactions between nodes. Links: regulation of gene expression by transcription factors
Hub: a highly connected node. Hub: p53 protein in cell cycle
Degree or connectivity: number of links per node
Network
dynamics8,43,83
Changes in network properties over time. Deletion of a gene, emergence of new motifs or
modules by mutations
Gene regulatory
networks1,42,43,83
Collection of DNA segments in a cell that interact with each other through their RNA and
protein expression products, thereby governing the rates at which genes in the network are
transcribed into messenger RNA. Interferon regulatory network
Pathway77,83
Metabolic pathway: a sequence of chemical reactions in a cell undergone by a group of molecules
leading to a predicted functional outcome. e.g., Amyloid pathway in Alzheimer’s disease
pathogenesis
Neural pathway: a neural tract connecting one part of the nervous system with another. e.g.,
Dopaminergic pathways: neural pathways in the brain that transmit the neurotransmitter
dopamine
Pathway
modularity77,83
A module is a self-contained component of a system, which has a well-defined interface to the
other components. e.g., Metabolic modules in the cells: cell cycle, glycolysis, energetic function
(mitochondria), receptor-signaling system
Pathway
analysis77,83
Study of a biochemical pathway instead of single genes or molecules. e.g., Axon guidance
pathway, amyloid pathway, oxidative stress pathway
Dynamic
disease10,54,55
A disease whose pathogenesis is mainly caused by the appearance of new dynamics of the
organism behavior, independently of the underlying pathogenesis. e.g., Several brain injuries
(tumor, trauma, and so forth) may affect neural networks generating seizures; Parkinson’s disease
symptoms are the results of disturbance of cortical-subcortical neural networks; epilepsy and
movement disorders are dynamic diseases
Neuromodulatory
system72,73
The neural network that modulates the overall activity of the excitatory (glutamatergic) and
inhibitory (GABAergic) networks. e.g., Dopaminergic, cholinergic, neural peptides neural
networks
Systems biomarker
discovery92
Identification of new biomarkers based in systems properties of the molecules involved. e.g.,
Ongoing
Systems drug
discovery83,84,91
Development of new therapies using systems biology analysis and/or aimed to modifying the
system properties of the tissue involved (in comparison with target-centric drug discovery). e.g.,
Ongoing
Bottom-up
approach1,3
Piecing together systems to create grander systems to gain insight into its emergent properties.
e.g., Identifying molecules and pathways implicated in the pathogenesis of neurological diseases,
providing an integrated model of the pathogenesis
Top-down
approach3,55
Breaking down a system to gain insight into its compositional subsystems. e.g., Starting from the
observable phenomena (cognition), identifying the domains and neuronal networks involved
through electroencephalogram, functional magnetic resonance imaging, or cognitive studies and
developing models of it
Villoslada et al: Systems Biology in Neurology 127
ease (AD) has been quite productively focused on the
identification of genetic and molecular perturbations in
the amyloid-␤ and -␶ pathways.24
However, the presence
of genetic heterogeneity and lack of knowledge
about the relative contribution of each defect to the
overall phenotype has delayed progress in the development
of therapeutic approaches.25
Recent studies demonstrated
that the pathways affected in individuals with
familial and sporadic AD are not the same. Furthermore,
they also highlighted the contribution of additional
processes such as axonal pathology,26,27
myelina-
tion,28
neuroinflammation,29
ischemic insult,30
and
aging31
to the overall pathogenesis. Intriguingly, it is
not uncommon to observe the opposite phenomenon,
in which alterations in the same pathway (amyloid)
lead to different disease phenotypes such as AD versus
inclusion body myositis. It has recently been shown
that computational models can be used to analyze perturbations
in multiple pathways over long periods that
may be responsible for the observed neuronal damage
in AD.31
Indeed, the strong correlation between the
development of neurodegenerative diseases and aging
suggests that even minor imbalances in these pathways
sustained for several years can finally overpower innate
repair mechanisms and result in the neurodegenerative
trait known as AD.32
Many of the genes and pathways altered in one neurological
disease are also commonly dysregulated in
others, even if they do not share the same pathophysiology,
age of onset, or outcome. Some diseaseactivated
pathways may represent an attempt by the
brain to restore homeostasis. Examples of this behavior
are the induction of small heat shock proteins with a
guardian-like protective role in the brain such as ␣␤crystallin,
the production of restorative secreted factors
such as neurotrophic factors, and the activation of mi-
croglia.33–35
Physiological systems, including the central
nervous system, may possess a limited ability to
deal with genetic or environmental perturbations, so
that different insults (eg, trauma, ischemia, metabolic
imbalances) trigger the same potentially restorative
pathways. Depending on genetic predisposition and
environmental exposures, a given individual will go on
to experience development of one of a diverse array of
conditions, such as Parkinson’s disease (PD), dementia,
leukodystrophies, and multiple sclerosis (MS), all of
which share common pathological processes.
Indeed, a recent large-scale analysis of brain transcripts
from neuropathological specimens demonstrated
unexpected similarities between MS and the metabolic
diseases hexosaminidases. These include the shared expression
of a constellation of inflammatory markers
such as cytokines, heat shock proteins, and other stress
proteins. Some of the most abundant transcripts found
in the affected brains from MS, Tay–Sachs, and Sandhoff’s
disease cases, but not in normal control cases,
include osteopontin, ␣␤-crystallin, major histocompatibility
class II, prostaglandin D2, calcyclin, apolipoprotein
E, metallothionein, dnaJ, monoamine oxidase B,
S-100, and calponin.36–38
Notably, these results were
not seen in sphingolipidoses. These data suggest that
there is an unexpected inflammatory component in
biochemical disorders of the brain. Interestingly, these
findings may have therapeutic implications. When
mice with deletions in the immunoglobulin Fc receptor
are bred with mice carrying the Tay–Sachs and Sandhoff’s
mutations, there is amelioration of disease.39
Fc
receptors are present in MS lesions, and deletion of Fc
receptors also ameliorates experimental autoimmune
encephalomyelitis, an animal model of MS.36
These
data suggest that two inherited biochemical disorders
of neural development can be modulated by regulating
pathways usually considered to be solely inflammatory.
This again illustrates that common mechanism or response
to injury can be activated in response to different
insults or pathogenic mechanisms. Moreover, such
commonalities can be better understood from an integrative
view of the many brain pathways at work during
central nervous system injury.
Neurological Diseases Can Be Studied in the
Context of Complex Networks
Among other approaches, systems biology uses
network-based analyses as a strategy for integration of
data from genetic, gene expression, proteomic, and
neurobiological experiments with the ultimate goal of
identifying pathways involved in the pathogenesis of
neurological diseases. The representation of real-world
systems by the network analogy represents another example
of inductive modeling that helps scientists approach
the complexity of the problem in question.
Networks are systems of interconnected entities. When
networks are studied in aggregate, certain properties
emerge from them that cannot be derived from the individual
analysis of each of their components. In graph
theory, networks are defined by nodes (representing
genes, proteins, or cells) and links, representing the interaction
among nodes. In homeostasis, biological networks
are constantly challenged and subjected to internal
or external perturbations (ie, via gene mutations or
environmental changes) that are normally counterbalanced
by the network’s robustness and adaptability (ie,
networks maintain their properties and continue to
perform their functions). However, if perturbations affect
key components of a network or are sustained for
a long period, a qualitative change occurs such that the
whole system is pushed toward a different (in this case,
pathological) steady state or “disease network.”
There are two main approaches for studying networks:
(1) structural, also referred to as topological analysis,
which examines the architecture of the system
(network connectivity patterns); and (2) dynamic anal-
128 Annals of Neurology Vol 65 No 2 February 2009
ysis, which examines how networks evolve over time by
changes in the number of nodes and in their connections.
Topological analysis is based on the statistical
properties of the network, such as the degree of connectivity,
mean path length, or clustering coefficient
(see Barabasi and Oltvai’s8
review). Many real networks
display the “small-world” property, in which several
nodes are locally connected among each other but
loosely connected to others, creating a modular structure.
Recent studies have demonstrated that, in smallworld
networks, it is possible to navigate from one particular
region to any other in just a few steps (eg, the
well-known phenomenon of “six degrees of separation”
among all people in the world7
). In scale-free networks,
the majority of nodes are poorly connected with other
nodes, but a few nodes (hubs) are highly connected (ie,
Google is an example of a hub in the Internet because
millions of other pages are just one click away). One of
the most salient properties of scale-free networks is that
they are robust against random attacks and sensitive to
selective targeting (ie, a single directed attack at Yahoo
will cause more damage to the Internet than will several
thousand random attacks at home PCs). These
properties are good examples of the “general solutions”
that can be achieved by deductive analysis of data from
several unrelated systems. Although these network
properties are characteristic of hard-wired networks, biological
networks are less fixed. Thus, considerable efforts
are now under way to characterize the dynamics
of biological networks.
Networks and Systems Control
When networks are studied at the local level, other interesting
topological elements, such as modules and
motifs, emerge. A module is a set of nodes participating
in similar functions that cluster together (ie, there
are more connections among themselves than there are
with any of the other molecules in the network).40,41
Examples of modules are abundant in analyzing key
biological functions. Many apoptotic events are centered
on caspases, whereas the cell cycle is coordinated
via cyclins. At the center of many key processes associated
with inflammation, the nuclear factor-␬B pathway
commands a huge position of importance.
Caspases in apoptosis, cyclins in the cell cycle, and nuclear
factor-␬B represent some examples of these socalled
local structures.
Network motifs are responsible for many basic control
mechanisms such as negative or positive feedback
and feed-forward signaling (Fig 1). The study of local
motifs integrates the topological and dynamic features
of complex networks.42,43
Feedback loops are an efficient
mechanism applied by cells to ensure precise control
over gene and protein expression, even in the presence
of noise from other cellular components. A
positive feedback enables the generation of bi(multi)stable
responses depending on the duration and intensity
of the initial stimulus. Bistability is the proposed
mechanism behind the storage of information and cellular
decisions (ie, cell phenotype).42,44
Bistability is
the existence of two stable states, and it is generated by
an abrupt transition in the dynamics of a system that
generates the new state (ie, activation of a gene expression
pattern leading to a permanent cell phenotype). A
negative feedback loop is an efficient mechanism for
maintaining product levels in a tightly regulated range,
and it is one of the most basic cellular mechanisms for
controlling homeostasis.45
Oscillatory dynamics is
common in signaling pathways, such as nuclear
factor-␬B pathway. More complicated dynamics can
appear with combinations of complex network motifs
that include positive and negative feedback loops. Finally,
feed-forward motifs provide a redundant mechanism
for the transmission of molecular information by
extending the duration of the signal and ensuring its
arrival at the intended site. Recent studies have shown
that these and other basic control motifs are present in
large regulatory networks in yeast, bacteria, and humans,
and form the basis for information processing
through cells.42,43
The abstract representation of complex
biological relationships using engineering tools
provides another example of the widespread use of inductive
models in systems biology. By mainly focusing
on certain aspects of the real world, these models can
provide important clues about how these systems work,
although it will take time before we know the degree to
which that conceptual representation provides insight
into the actual mechanisms.
Like engineering-based systems-control analysis, the
study of complex sets of biological entities in the context
of networks may demonstrate common properties
relevant for disease pathogenesis that cannot be predicted
by analyzing each entity in isolation. Several
studies have questioned whether proteins associated
with diseases share any particular network property,
such as higher connectivity, preferential modular architecture,
or involvement in more than one biological
pathway.46–49
Although still in their infancy, these approaches
may provide important insights toward understanding
the pathogenesis of neurological diseases
and designing new therapeutic strategies. For example,
if disease-related proteins interact with several other
proteins (are hubs) within their networks (eg, p53 in
cancer), any pharmacologically induced modulation
will affect the overall proteome network and may induce
significant side effects. Moreover, by relating
mechanism of action and side effects through network
analysis, new applications of existing drugs may be in-
ferred.50
A recent study analyzing the protein network involved
in inherited ataxias exemplifies the power of this
approach.49
In this study, Lim and colleagues49
examVilloslada
et al: Systems Biology in Neurology 129
ined the biological relations between the mutated proteins
in hereditary ataxias. They hypothesized that if
mutations in different genes lead to death of Purkinje
neurons in all ataxias, they should be part of common
biological pathways critical for the survival of these cerebellar
neurons. To answer that question, they performed
a topological network analysis of the protein–
protein interaction network. Starting from the 23
proteins previously known to be involved in hereditary
ataxias because of the presence of known mutations in
their genes, they identified the contextual network of
3,607 proteins in which the ataxia network was included.
The gene ontology analysis showed that proteins
affected in ataxias are functionally clustered in related
pathways (Fig 2A), with overrepresentation of
pathways related to RNA splicing, ubiquitination, and
the cell cycle. This finding suggests that RNA splicing
plays a critical role in Purkinje cell degeneration. By
analyzing the ataxia network, Lim and colleagues49
were able to confirm involvement of the previously described
pathways and the recently discovered
Puratrophin-1 as an ataxia gene51
having strong interactions
with Ataxin-1 (see Fig 2B), even though
Puratrophin-1 was not included in the original list of
ataxia genes. Overall, the protein network analysis of
human ataxias provided an integrated view of these diseases
as RNA splicing diseases promoting the death of
Purkinje neurons. This aspect of systems biology will
open new avenues for the development of new diagnostic
tests and therapies.
In another study, Miller and coauthors31
applied a
network analysis to DNA array studies from the CA1
region of the hippocampus of patients with AD and
normal aging individuals. Instead of reporting a list of
differentially expressed genes between both conditions,
an approach with low probability of replication, they
used weighted gene coexpression network analysis to
group results into a nine functionally relevant pathways.
They found that synaptic transmission, extracellular
transport, immune response, mitochondrial and
metabolic processes, and myelination were associated
with disease progression. Some of these modules correlate
with the degree of cognitive impairment measured
with the Mini-Mental test or with the pathological
burden measured using neurofibrillary tangle quantification.
Moreover, a commonality between AD and aging
was found in the involvement of mitochondrial
processes and synaptic plasticity. In addition, Miller
and coauthors31
found new evidence supporting the
role of demyelination and oligodendrocyte dysfunction
in AD progression, which was related to the
preselinin-1 pathway. Overall, by providing a pathway
view of AD pathogenesis, these authors provided a new
set of mechanistic hypotheses about AD that can be
tested in future studies. The benefits of these kinds of
approaches would be exemplified in that they help in
Fig 1. Genetic circuits. (A) In positive feedback loop (left), the
expression of a gene product is stimulated by its own expression,
thus enabling the generation of bistable responses (right)
depending on the duration and intensity of the initial stimulus.
(B) Another basic type of control is negative feedback
loops (left). When levels of gene A increase, expression of gene
B is induced, which represses expression of gene A (and its
own expression). When levels of A fall below a threshold, its
production is stimulated because of the absence of the repressor
B. This control mechanism yields oscillations around a mean
value (right), which, in turn, depends on other parameters of
the system such as synthesis and degradation rates. (C) Feedforward
loops (FFL) consist of three genes: A, B, and C. Gene
A is a regulator of B and C (left). FFL is coherent if the sign
(activation or suppression) of the path A-B-C is the same as
the sign of the path A-C. If the signs do not match, the FFL
is incoherent. FFL can reject transient inputs and activate
only after persistent stimulation (ie, TrkA and calcium signaling,
right). (D) The bifan motif (left), composed of two source
nodes directly cross-regulating two target nodes, is able to act
as signal sorter, a synchronizer, or a filter (ie, glycine pathway,
right). It also provides temporal regulation of signal
propagation. (E) In the single-input-module motif (SIM),
modulator A controls the expression of a group of species (left).
This motif can generate an ordered expression program for
each of the components under regulation of A (ie, rtg1mitochondrial
pathway in yeast, right). mRNA ϭ messenger
RNA; NMDAR ϭ N-methyl-D-aspartate receptor; PI3K ϭ
phosphatidylinositol 3 kinase; PKC ϭ protein kinase C.
130 Annals of Neurology Vol 65 No 2 February 2009
the design of new therapies targeting the pathways in-
volved.
In addition to studying ensembles of genes or proteins,
network analysis offers a potentially useful way to
improve the classification of neurological diseases by
closing the gap between causative factor, pathogenesis,
and disease phenotypes.52
Recently, Goh and research-
ers47
analyzed data from the Online Mendelian Inheritance
in Man (OMIM) database and characterized the
full set of disease–gene associations in what they called
the “human diseasome” (Fig 3). In this analysis, two
genes are linked together if they are responsible for
causing the same disease, and two diseases are linked
together if they share at least one defective gene. The
authors found that the products of genes associated
with phenotypically similar disorders tend to cluster together.
Indeed, experiments demonstrated that proteins
affected by the same disease interact in such networks
more frequently with each other than with those not
associated with the disease. These results support the
concept of disease-specific functional modules. Analysis
of the components of networks may show certain common
responses to brain pathology and offers a new formalized
conceptual approach to assess disease susceptibility.
Thus, network analysis and other systems
approaches can be helpful for neurologists in the process
of classifying diseases.
We must clarify that the methods identified earlier
are used to find patterns in data. Depending on the
nature of the data and the method used, the pattern
found is mapped formally in a geometric point-topoint
manner to an abstract, idealized pathway or network.
As with any other model, the pathways and networks
described here are inductive concepts, and no
quantitative mapping is possible between the conceptual
system and the real system. Nevertheless, biomedical
research relies heavily on conceptual models, and
the formal introduction of these should come as no
surprise to the reader. An important contribution of
these models is that they provide the guidelines needed
to construct analogue mechanisms (a task that remains
to be done) using the irreplaceable methods of experimental
biology. The map between those mechanisms
and biology can be iterated in a concrete manner.53
Dynamic Analysis of Neurological Disease
Chronic diseases can be distinguished by their onset
(acute, subacute, or progressive) and their subsequent
Fig 2. Ataxia network. By performing protein interaction studies, bioinformatics search, and network analysis, Lim and colleagues49
hypothesized the existence of a network, pathway modules, and functions that may relate the genes associated with human hereditary
ataxias. (A) Enriched gene ontology (GO) categories of the ataxia network. Pie charts depict the relative number of GO terms enriched
in each of the three branches of the GO structure when using the whole genome and/or the hORFeome as reference distributions.
Enriched categories are identified as those significant (p Ͻ 0.05) after adjusting for multiple testing. Bar charts depict the
enriched GO terms from the common set (whole genome and hORFeome). The number of bars shown is less than the total number
of common enriched categories because bars represent only the terminal GO terms. The length of the bar is the standardized enrichment
score. (B) Graphic representation of an ataxia subnetwork. Several interactors of ATXN1 appeared to be genetic modifiers as
well. Nodes in yellow represent interacting proteins, hypothesized ataxia-related proteins are in blue, and putative genetic modifiers
of SCA1 pathology are in orange. Squares and circles indicate two-hybrid baits or prey, respectively. Edge colors represent the different
hypothetical interactions: green (from this study) and blue literature curated interactions (LCI). (Reprinted from Lim and
colleagues,49
by permission.)
Villoslada et al: Systems Biology in Neurology 131
Fig 3. Human disease network. By relating diseases with gene mutations available at OMIM, Goh and researchers47
built a network
in which diseases were grouped based on their genetic commonalities. (A) The human disease network: Each node corresponds
to a specific disorder colored by class (22 classes). The size of each node is proportional to the number of genes thought to contribute
to the disorder. Edges between disorders in the same disorder class are colored with the same (lighter) color, and edges connecting
different disorder classes are colored gray, with the thickness of the edge proportional to the number of genes shared by the disorders
connected by it. (B) The disease gene network: Each node is a single gene, and any two genes are connected if implicated in the
same disorder. In this network map, the size of each node is proportional to the number of specific disorders in which the gene has
been implicated. The resulting network challenges current disease classifications based on phenotypes (syndromes), with implications
for diagnosis and therapeutics. (Reprinted from Goh and researchers,47
by permission.)
132 Annals of Neurology Vol 65 No 2 February 2009
clinical course (self-limiting, relapsing remitting, cyclic,
or chronic progressive). Recognition of the critical
value of time and temporal rhythms, combined with
the knowledge that particular temporal rhythms respond
best to certain treatment strategies, often provides
a basis for therapy. Chronic diseases develop over
variable but extended periods until a phenotype manifests.
This time dependency suggests that the normal
physiological processes are robust but subject to failure,
possibly because of a number of mechanisms (eg, redundancy,
network topology). Only the prolonged exposure
to pathological insults is capable of shifting the
system to another alternative robust, but unfortunately
pathological state. Under this scenario, complex diseases
arise because of abnormalities in the underlying
physiological control mechanisms, more so than in any
perturbation of one or a limited number of specific
molecular events.10,54–57
Thus, development of therapeutic
strategies aimed at reestablishing the altered dynamics
may be another rational approach toward cures
of diseases.58
The concept of dynamic diseases is deeply rooted in
neurology.58,59
The first descriptions of dynamic diseases
came from the phenomenology of seizures, paroxysmal
movement disorders, and psychiatric diseases
such as bipolar disorders.60–63
A classic example is the
development of seizures, which can be identified by
changes in the normal dynamics of the electroencephalogram
(EEG). Epileptic foci, which can be generated
by many different kinds of neuronal insults, generate a
new rhythm that spreads through brain networks, creating
a pathological brain dynamic.60
Movement disorders,
such as tremor or gait abnormalities in PD or
Huntington disease, are also examples of alterations in
the dynamics of physiological systems.61
The motor
symptoms of PD have been associated with increased
activity of the subthalamic nucleus, resulting in excessive
inhibitory outflow from the basal ganglia to the
thalamus and brainstem.64
High-frequency deep brain
stimulation modulates the activity of the subthalamic
nucleus and restores normal dynamics of the motor cir-
cuit.65
Dopaminergic therapy also suppresses the activity
of the subthalamic nucleus, modulating the dynamics
of the corticosubcortical system.66
The longitudinal analysis of the cellular network in
the basal ganglia in patients with PD using positron
emission tomography and analyzing the spatial covariance
in the data has identified a motor pattern characterized
by increased pallidothalamic and pontine metabolic
activity. This increased activity is associated with
reduction of premotor and posterior parietal cortical
regions (Fig 4).67
Disease progression was associated
with increasing metabolism in the subthalamic nucleus,
internal globus pallidus, dorsal pons, and primary motor
cortex.68
This analysis demonstrated that cognitive
impairment in PD patients was associated with a different
cellular network, with reduced metabolic activity
in the prefrontal and parietal cortices, and relative increases
in the dentate nuclei and cerebellar hemi-
spheres.69
Assessment of the effect of deep brain stimulation
and dopamine therapy in the dynamics of the
basal ganglia network70
demonstrated that both therapies
produced metabolic reductions in the precise areas
that are chronically stimulated in PD by the subthalamic
nucleus. These target areas for the subthalamic
nucleus included the putamen-globus pallidus, cerebellar
vermis, and parietal cortex. Although both approaches
showed different effectiveness in individual
regions, the overall activity of the network was similar,
and both approaches correlated well with clinical improvement.
Thus, such studies highlight the critical
role of brain dynamics in the development of neurological
diseases, which would not be unraveled through
dissection of the different components, and how such
dynamics may be rationally targeted through therapy.
Several methods allow for the monitoring of brain
dynamics, including EEG, positron emission tomography,
and functional magnetic resonance imaging.
These are examples of descriptive models whereby particular
aspects of brain function can be measured. Pattern
recognition methods are then applied to reconcile
the obtained data, and the inductive model investigators
then have insight about how that part of the brain
works. Methods such as nonlinear dynamic analysis,
computational models, or network analysis can be useful
for achieving such reconciliation between data and
model. For example, although slow periodic EEG discharges
are common in encephalopathies, encephalitis,
and tumors, their underlying pathogenesis remains unknown,
thus preventing the development of new therapies.
Frohlich and colleagues71
performed a dynamic
analysis of a computational model of the human cortex
to better understand the generation of slow periodic
EEG discharges. In particular, they examined the deafferentation
caused by neuronal loss. To simulate the
deafferentation process, they used data from
Creutzfeldt–Jacob disease and tested different degrees
of deafferentation for modeling different degrees of
neuronal loss and disease severity. They found that different
degrees of deafferentation influence the neural
network dynamics, with a critical degree (80%) of
deafferentation after which a new periodic dynamic of
slow waves emerge (Fig 5). This resulted in poor information
transmission through the neuronal network.
These unexpected results are a paradoxical effect of
brain plasticity. Thus, this modeling approach benefits
the understanding of encephalopathies because it provides
information about how much neural networks
can cope with brain damage and why this new slow
dynamics emerge. Moreover, such findings may lead to
the development of new therapies for preventing secVilloslada
et al: Systems Biology in Neurology 133
ondary damage caused by brain plasticity in neurological
diseases.
It is well known that the brain is organized in dynamic
neuronal networks of excitatory (glutamatergic)
and inhibitory (GABAergic) connections. The precise
activity of neuronal networks is modulated by several
regulatory pathways, such as dopaminergic, cholinergic,
serotoninergic, or histaminergic circuits, collectively
known as the neuromodulatory system.72–74
Unlike epilepsy,
in which excitatory-inhibitory networks are the
main pathways affected, most neurodegenerative diseases
show disruption of the neuromodulatory system (dopaminergic
system in PD, cholinergic system in AD). This
provides another example of emergent network properties.
In this case, although the loss of a significant but
random number of neurons throughout the brain may
not have a tangible clinical impact, when an insult targets
a critical part of the network such as the dopaminergic
system in PD, it starts a cascade of failures that
ultimately culminates in progressive neurological sympFig
4. Parkinson’s disease (PD)–related spatial covariance pattern. (A) Parkinson’s disease–related spatial covariance pattern
(PDRP) identified by network analysis of 18F-fluorodeoxyglucose (FDG) positron emission tomographic scans. This spatial covariance
pattern was characterized by relative metabolic increases in the putamen, globus pallidus (GP), thalamus, pons, and cerebellum,
and was associated with decreases in the premotor and posterior parietal areas. The display shows voxels that contribute significantly
to the network (voxels with positive region weights [metabolic increases] are color coded from red to yellow; voxels with
negative region weights [metabolic decreases] are color coded from blue to purple). (B) Region weights on PDRPs seen in the rest
state. The patterns were characterized by substantial contributions from the putamen or GP, thalamus, and cerebellum (metabolic
increases), and from the lateral premotor and parietal association regions (metabolic decreases). Black line indicates populationaveraged
regional loadings across centers. (C) Network activity was increased in PD. (D) Mean PDRP activity in patients with
early-stage PD followed up longitudinally at baseline, 24 months, and 48 months. Network activity increased over time. PDRP
expression in the patient group was significantly increased at all three time points relative to values for healthy control subjects.
Dashed line is one standard deviation (SD) above the normal mean. Bars show the standard error for the PD patient group at
each time point. Asterisks show the significance of comparisons with control values at each time point. (Reprinted from Eckert and
colleagues,67
by permission.)
134 Annals of Neurology Vol 65 No 2 February 2009
toms. Damage of this neuromodulatory system also illustrates
the concept of dynamic diseases, because dopaminergic
neuronal loss alters the dynamics of basal
ganglia cortical circuits as described earlier, ultimately
leading to clinical symptoms.
Pathway-Centered Analysis of Neurological
Diseases
Pathway analysis is becoming a valuable tool for improving
our understanding of molecular data generated
from human studies. The studies of chemical pathways
can reconcile many studies, providing valuable data but
not confirming one to the other because pathway analysis
is focused in the outcome of the system. By this
way, the study of pathways buffers the heterogeneity
because individual patients bear mutations or damage
in different molecules of the same pathway leading to
the same outcome (ie, protein misfolding in neurodegenerative
diseases). For example, studies of hereditary
PD have identified six genes involved in its pathogenesis:
SNCA (␣-synuclein), PARK2 (parkin), DJ-1,
UCHL1, PINK1, and LARK2 (dardarin).75,76
The clinical
and pathological heterogeneity of individuals with
such mutations has prompted researchers to redefine
PD and its nosology. The functions of genes identified
in hereditary PD involve lipid and vesicle dynamics (␣synuclein),
the ubiquitin-proteasome system (parkin
and UCHL1), mitogen-activated protein kinase kinase
kinase (MAPKKK) signaling leucine-rich repeat kinase
2 (LRRK2), oxidative stress and mitochondrial function
(DJ-1, PINK1, parkin), and microtubule stability.
However, the lack of a comprehensive view of how
these pathways interact to cause brain damage has
hampered our understanding of the pathogenesis of
PD.76
One strategy involves the use of high-throughput approaches
such as large-scale DNA microarrays and proteomics
in pathological specimens. The confluence of
such studies demonstrates the remarkable interplay of
genes regulating common pathways in a disease, a concept
now termed genomic convergence.77,78
Using this
approach, Hauser and investigators77
were able to map
genes differentially expressed in the substantia nigra of
control subjects and patients suffering from PD to five
genomic loci, and they identified the mitochondrial
pathways as the main candidates for the susceptibility
for PD. Another example of pathway analysis and
genomic convergence in PD is the integration of genetic
and gene expression studies to support the involvement
of the axon-guidance pathways (ephrins, netrins,
semaphorins, slits and their receptors, and
intermediate proteins).79,80
The study of PD and other
degenerative diseases using pathway analysis can help
to integrate the biological data in functional models
(pathways), coping with disease heterogeneity and providing
therapeutic targets for treating such disease. In
the aforementioned example of axon-guidance pathways,
their involvement contributes to improving our
understanding of the pathogenesis of PD with another
biological function (in addition of dopamine dysfunction
and oxidative stress) not previously recognized and
because it provides new targets for treating PD.
Guiding Drug Discovery and Personalized
Therapy
Neurology is now entering a new therapeutic age. Instead
of symptomatic therapies, neurologists are developing
therapies that modify disease pathogenesis. This
will impose several challenges because of the critical
importance of neural dynamics in the generation of
neurological symptoms. For example, full blockade of
lymphocyte homing to the central nervous system in
MS led to the undesired development of progressive
multifocal encephalopathy in patients treated with na-
talizumab.81
Similarly, vaccination of AD patients with
␤ amyloid led to the development of autoimmune encephalitis
in 10% of the patients in a clinical trial.82
Targeting molecules without a thorough understanding
of the integrated mechanisms in which they participate
often leads to unexpected and sometimes catastrophic
adverse effects.83
The approach taken by systems biology involves a
different paradigm based on improved knowledge of
dynamic mechanisms.20,84,85
For example, the in vivo
Fig 5. Computational analysis of periodic slow-wave patterns
in neurological diseases. (A) Activity map of pyramidal neurons
(PYs; top) and interneurons (INs; bottom). Cool and hot
colors indicate hyperpolarization and depolarization, respectively.
Severe deafferentation (90%) induced consecutive decline
in network activity (arrowhead). Recovery of target firing
rate by homeostatic scaling resulted in prominent periodic
network activation. (B) Simulated local field potentials (lfp):
Before deafferentation, lfp have high-frequency activity with
low amplitude. After deafferentation (arrowhead), the recovery
of activity level is characterized by slow high-amplitude lfp
oscillations. (Reprinted from Frohlich and colleagues,71
by
permission.)
Villoslada et al: Systems Biology in Neurology 135
transcriptional response to interferon-␤ was recently
studied using a network-based conceptual analysis.86
This study demonstrated that, in addition to reducing
T-cell adhesion, interferon-␤ represses RNA transcription
and protein synthesis, and induces apoptosis in
these cells. These findings suggest that the most widely
used therapy to control exacerbations in MS may
achieve its benefits by simultaneously targeting components
of interacting and intertwined mechanisms.
By targeting key steps in the dynamics of a disease at
different time points and targeting different regions or
modules of the molecular or cellular mechanisms governing
the pathogenesis of the disease, better regulation
of the system under study can be achieved. For example,
a recent analysis of the gene expression network
that controls T-cell activation in patients with MS allowed
identification of Jagged-1 as a new therapeutic
target.85
The reductionist approach would have prevented
such discovery, because levels of Jagged-1 did
not differ between patients and control subjects, but
network analysis identified the critical role of this gene
in the set of interactions regulating T-cell differentiation
in MS patients. Moreover, in vitro assays and network
analysis identified a synergistic effect between
Jagged-1 and interferon-␤, indicating that systems drug
discovery may improve our ability to identify new therapies
for neurological diseases.
Computational modeling can be helpful in the design
of the most adequate strategies in this multitarget
and multistep process, as well as in the prediction of
some unexpected side effects.87–89
For example, mathematical
modeling of the multifactorial events in brain
damage may allow the identification of optimal therapeutic
strategies, as recently demonstrated in spinal
cord injury.90
Indeed, the integrative approach of systems
biology may speed up and reduce the costs of the
current drug discovery process by testing combination
therapy, and identifying the best targets and dynamics
to be modified using computational models and validated
cell assays.84,91
A crucial concept is that the suitability
of a given protein as therapeutic target is determined
by the nature of its contribution to the
network’s control mechanisms (pathway drug target)
rather than by its aberrant activity or expression (protein
drug target).83
These approaches can also be helpful for the identification
of new biomarkers,92
a critical step in the process
of developing stratified and personalized medi-
cine.93
For example, network analysis of prostatic
cancer identified the androgen receptor as a novel
pathway and a genetic mediator of metastasis, suggesting
that targeting of this pathway may rescue patients
escaping androgen dependence.94
Also, the analysis of
serum proteome from patients with AD demonstrated
the presence of abnormalities in the blood, indicating a
deregulation of hematopoiesis, immune responses, apoptosis,
and neuronal support.95
Proteomic studies of
MS lesions have demonstrated numerous unexpected
pathways in various pathological stages of disease.34
For example, elements of the coagulation cascade, including
tissue factor and the inhibitor of protein C,
were shown to be critical in active and chronic active
MS lesions. In models of MS, paralysis could be reversed
with activated protein C, an approved drug for
treatment of septic shock, and with thrombin inhibitors
used as approved anticoagulants.34
Finally, a recent
report suggested that transcriptional profiling of
CD4 T cells in clinically isolated syndrome patients accurately
predicts their conversion to clinically definite
MS.96
Together, these findings suggest that genetic
and biochemical abnormalities involved in the pathogenesis
of neurological diseases may affect other systems
without inducing pathological effects.
Concluding Remarks
It is imperative that any scientific research, especially if
it requires an interdisciplinary approach, be grounded
in the philosophy of science and furnished with the
logical tools that permit the translation of empirical
data into useful knowledge and worthwhile means of
advancement.97
The knowledge generated in the past
few decades on the pathogenesis of neurological diseases,
as well as the development of high-throughput
methods of analysis, biotechnology, and computational
biology, has provided unprecedented opportunities for
developing new disease-modifying therapies. We argue
that the integration of this knowledge into the theoretical
framework and the tools provided by systems biology
will be an invaluable help in this process. However,
therapeutic decisions made by neurologists as we
treat patients will always be based on the expertise of
the physician and on evidence-based medical practice.
The genomic portrait of an individual may allow a predictive
and personalized approach to therapy. In the
near future, this process will be aided by powerful
computers, with information technologies that will
manage the available information from the patient’s
tests, the patient’s medical history and genetic history,
and ever-evolving scientific databases.21
Although systems
biology offers a new framework for the study of
human disease, better characterization of biological
processes, quantitative data, and dynamic information
are still required to successfully translate this paradigm
from the bench/chip to the bedside. However, we find
ourselves at an exciting turning point in the history of
biomedical research, where integrated methods for the
analysis of complex diseases allowing the conversion of
information into effective knowledge are finally within
our grasp.
136 Annals of Neurology Vol 65 No 2 February 2009
Note Added in Proof
Another recent review regarding neurodegenerative diseases
was recently released: Deciphering complex
mechanisms in neurodegenerative diseases: the advent
of systems biology. Noorbakhsh F, Overall CM, Power
C. Trends Neurosci 2009 Jan 8.
This work was supported by the Pathfinder Initiative on Complexity
for Living Science of the European Union (ComplexDis [NEST-
043241], P.V.).
Acknowledgment
We also thank C. Anthony Hunt for critical reading of
this manuscript. Pathway examples for each motif were
obtained from Ravy Lyengar Lab Web site at Mount
Sinai (http://www.mssm.edu/labs/iyengar/resources)
and from Richard Young Lab at Massachusetts Institute
of Technology (http://jura.wi.mit.edu/cgi-bin/
young_public/
References
1. Ideker T, Galitski T, Hood L. A new approach to decoding life:
systems biology. Annu Rev Genomics Hum Genet 2001;2:
343–372.
2. Kitano H. Systems biology: a brief overview. Science 2002;295:
1662–1664.
3. Rosenblueth A, Wiener N. The role of models in science. Philos
Sci 1945;12:316–321.
4. Kola I, Landis J. Can the pharmaceutical industry reduce attrition
rates? Nat Rev Drug Discov 2004;3:711–715.
5. Baranzini SE. Systems-based medicine approaches to understand
and treat complex diseases: the example of multiple sclerosis.
Autoimmunity 2006;39:651–662.
6. Germain RN. The art of the probable: system control in the
adaptive immune system. Science 2001;293:240–245.
7. Watts DJ, Strogatz SH. Collective dynamics of ‘small-world’
networks. Nature 1998;393:440–442.
8. Barabasi AL, Oltvai ZN. Network biology: understanding the
cell’s functional organization. Nat Rev Genet 2004;5:101–113.
9. Bertalanffy LV. General system theory: foundations, development,
applications. New York: George Braziller, 1976.
10. Glass L. Synchronization and rhythmic processes in physiology.
Nature 2001;410:277–284.
11. Kitano H. Computational systems biology. Nature 2002;420:
206–210.
12. Joyce AR, Palsson BO. The model organism as a system: integrating
‘omics’ data sets. Nat Rev Mol Cell Biol 2006;7:
198–210.
13. Sauer U, Heinemann M, Zamboni N. Genetics. Getting closer
to the whole picture. Science 2007;316:550–551.
14. Wagner A, Fell DA. The small world inside large metabolic
networks. Proc Biol Sci 2001;268:1803–1810.
15. Wagner A. Robustness and evolvability in living systems.
Princeton, NJ: Princeton University Press, 2005.
16. Kitano H. Biological robustness. Nat Rev Genet 2004;5:
826–837.
17. Stelling J, Sauer U, Szallasi Z, et al. Robustness of cellular functions.
Cell 2004;118:675–685.
18. Wiener N. Cybernetics. New York: Wiley, 1948.
19. Cannon W. Organization for physiological homeostasis. Physiol
Rev 1929;9:397.
20. Kitano H. A robustness-based approach to systems-oriented
drug design. Nat Rev Drug Discov 2007;6:202–210.
21. Villoslada P, Oksenberg J. Neuroinformatics in clinical practice:
are computers going to help neurological patients and their
physicians? Future Neurol 2006;1:1–12.
22. Glimcher P. Decisions, uncertainty, and the brain: the science
of neuroeconomics. Cambridge, MA: MIT Press, 2004.
23. De Schutter E. Why are computational neuroscience and systems
biology so separate? PLoS Comput Biol 2008;4:e1000078.
24. Mattson MP. Pathways towards and away from Alzheimer’s disease.
Nature 2004;430:631–639.
25. Dragunow M. The adult human brain in preclinical drug development.
Nat Rev Drug Discov 2008;7:659–666.
26. Stokin GB, Lillo C, Falzone TL, et al. Axonopathy and transport
deficits early in the pathogenesis of Alzheimer’s disease.
Science 2005;307:1282–1288.
27. Adalbert R, Gilley J, Coleman MP. Abeta, tau and ApoE4 in
Alzheimer’s disease: the axonal connection. Trends Mol Med
2007;13:135–142.
28. Bartzokis G, Sultzer D, Lu PH, et al. Heterogeneous agerelated
breakdown of white matter structural integrity: implications
for cortical “disconnection” in aging and Alzheimer’s disease.
Neurobiol Aging 2004;25:843–851.
29. Wyss-Coray T. Inflammation in Alzheimer disease: driving
force, bystander or beneficial response? Nat Med 2006;12:
1005–1015.
30. Zlokovic BV. Neurovascular mechanisms of Alzheimer’s neurodegeneration.
Trends Neurosci 2005;28:202–208.
31. Miller JA, Oldham MC, Geschwind DH. A systems level analysis
of transcriptional changes in Alzheimer’s disease and normal
aging. J Neurosci 2008;28:1410–1420.
32. Palop JJ, Chin J, Mucke L. A network dysfunction perspective
on neurodegenerative diseases. Nature 2006;443:768–773.
33. Hauser SL, Oksenberg JR. The neurobiology of multiple
sclerosis: genes, inflammation, and neurodegeneration. Neuron
2006;52:61–76.
34. Han MH, Hwang S, Roy DB, et al. Proteomic analysis of active
multiple sclerosis lesions reveals therapeutic targets in the
coagulation cascade. Nature 2007;451:1076–1081.
35. Villoslada P, Genain CP. Role of nerve growth factor and other
trophic factors in brain inflammation. Prog Brain Res 2004;
146:403–414.
36. Lock C, Hermans G, Pedotti R, et al. Gene-microarray analysis
of multiple sclerosis lesions yields new targets validated in autoimmune
encephalomyelitis. Nat Med 2002;8:500–508.
37. Myerowitz R, Lawson D, Mizukami H, et al. Molecular pathophysiology
in Tay-Sachs and Sandhoff diseases as revealed by
gene expression profiling. Hum Mol Genet 2002;11:
1343–1350.
38. Myerowitz R, Mizukami H, Richardson KL, et al. Global gene
expression in a type 2 Gaucher disease brain. Mol Genet Metab
2004;83:288–296.
39. Yamaguchi A, Katsuyama K, Nagahama K, et al. Possible role
of autoantibodies in the pathophysiology of GM2 gangliosidoses.
J Clin Invest 2004;113:200–208.
40. Qi Y, Ge H. Modularity and dynamics of cellular networks.
PLoS Comput Biol 2006;2:1502–1510.
41. Sauro HM. Modularity defined. Mol Syst Biol 2008;4:
166–167.
42. Ma’ayan A, Jenkins SL, Neves S, et al. Formation of regulatory
patterns during signal propagation in a mammalian cellular network.
Science 2005;309:1078–1083.
Villoslada et al: Systems Biology in Neurology 137
43. Alon U. Network motifs: theory and experimental approaches.
Nat Rev Genet 2007;8:450–461.
44. Losick R, Desplan C. Stochasticity and cell fate science. 2008;
320:65–68.
45. Smolen P, Baxter DA, Byrne JH. Mathematical modeling of
gene networks. Neuron 2000;26:567–580.
46. Goni J, Esteban FJ, Velez N, et al. A computational analysis of
protein-protein interaction networks in neurodegenerative diseases.
BMC Syst Biol 2008;2:52.
47. Goh KI, Cusick ME, Valle D, et al. The human disease network.
Proc Natl Acad Sci USA 2007;104:8685–8690.
48. Lu X, Jain VV, Finn PW, Perkins DL. Hubs in biological interaction
networks exhibit low changes in expression in experimental
asthma. Mol Syst Biol 2007;3:98.
49. Lim J, Hao T, Shaw C, et al. A protein-protein interaction
network for human inherited ataxias and disorders of Purkinje
cell degeneration. Cell 2006;125:801–814.
50. Campillos M, Kuhn M, Gavin AC, et al. Drug target identification
using side-effect similarity. Science 2008;321:263–266.
51. Ishikawa K, Toru S, Tsunemi T, et al. An autosomal dominant
cerebellar ataxia linked to chromosome 16q22.1 is associated
with a single-nucleotide substitution in the 5Ј untranslated region
of the gene encoding a protein with spectrin repeat and
Rho guanine-nucleotide exchange-factor domains. Am J Hum
Genet 2005;77:280–296.
52. Loscalzo J, Kohane I, Barabasi AL. Human disease classification
in the postgenomic era: a complex systems approach to human
pathobiology. Mol Syst Biol 2007;3:124.
53. Hunt CA, Baranzini SE, Matthay MA, Park S. A framework
and mechanistically focused, in silico method for enabling rational
translational research. American Medical Informatics Association
(AMIA) Summit on Translational Bioinformatics;
2008; San Francisco, CA.
54. Mackey MC, Glass L. Oscillation and chaos in physiological
control systems. Science 1977;197:287–289.
55. Varela F, Lachaux JP, Rodriguez E, Martinerie J. The
brainweb: phase synchronization and large-scale integration.
Nat Rev Neurosci 2001;2:229–239.
56. McCormick DA. Are thalamocortical rhythms the Rosetta
Stone of a subset of neurological disorders? Nat Med 1999;5:
1349–1351.
57. Bar-Yam Y. Dynamics of complex systems. Reading, MA:
Addison-Wesley, 1997.
58. Milton J, Black D. Dynamic diseases in neurology and psychiatry.
Chaos 1995;5:8–13.
59. Milton JG, Longtin A, Beuter A, et al. Complex dynamics and
bifurcations in neurology. J Theor Biol 1989;138:129–147.
60. Lopes da Silva F, Blanes W, Kalitzin SN, et al. Epilepsies as
dynamical diseases of brain systems: basic models of the transition
between normal and epileptic activity. Epilepsia 2003;
44(suppl 12):72–83.
61. Hausdorff JM, Cudkowicz ME, Firtion R, et al. Gait variability
and basal ganglia disorders: stride-to-stride variations of gait cycle
timing in Parkinson’s disease and Huntington’s disease.
Mov Disord 1998;13:428–437.
62. Llinas RR, Ribary U, Jeanmonod D, et al. Thalamocortical
dysrhythmia: a neurological and neuropsychiatric syndrome
characterized by magnetoencephalography. Proc Natl Acad Sci
USA 1999;96:15222–15227.
63. Pezard L, Nandrino JL, Renault B, et al. Depression as a dynamical
disease. Biol Psychiatry 1996;39:991–999.
64. Vitek JL, Giroux M. Physiology of hypokinetic and hyperkinetic
movement disorders: model for dyskinesia. Ann Neurol
2000;47:S131–S140.
65. Olanow CW, Brin MF, Obeso JA. The role of deep brain stimulation
as a surgical treatment for Parkinson’s disease. Neurology
2000;55:S60–S66.
66. Schapira AH, Bezard E, Brotchie J, et al. Novel pharmacological
targets for the treatment of Parkinson’s disease. Nat Rev
Drug Discov 2006;5:845–854.
67. Eckert T, Tang C, Eidelberg D. Assessment of the progression
of Parkinson’s disease: a metabolic network approach. Lancet
Neurol 2007;6:926–932.
68. Huang C, Tang C, Feigin A, et al. Changes in network activity
with the progression of Parkinson’s disease. Brain 2007;130:
1834–1846.
69. Huang C, Mattis P, Tang C, et al. Metabolic brain networks
associated with cognitive function in Parkinson’s disease. Neuroimage
2007;34:714–723.
70. Asanuma K, Tang C, Ma Y, et al. Network modulation in the
treatment of Parkinson’s disease. Brain 2006;129:2667–2678.
71. Frohlich F, Bazhenov M, Sejnowski TJ. Pathological effect of
homeostatic synaptic scaling on network dynamics in diseases of
the cortex. J Neurosci 2008;28:1709–1720.
72. Yu AJ, Dayan P. Uncertainty, neuromodulation, and attention.
Neuron 2005;46:681–692.
73. Hasselmo ME. Neuromodulation: acetylcholine and memory
consolidation. Trends Cogn Sci 1999;3:351–359.
74. Hasselmo ME, Giocomo LM. Cholinergic modulation of cortical
function. J Mol Neurosci 2006;30:133–135.
75. Hardy J, Cai H, Cookson MR, et al. Genetics of Parkinson’s
disease and parkinsonism. Ann Neurol 2006;60:389–398.
76. Farrer MJ. Genetics of Parkinson disease: paradigm shifts and
future prospects. Nat Rev Genet 2006;7:306–318.
77. Hauser MA, Li YJ, Takeuchi S, et al. Genomic convergence:
identifying candidate genes for Parkinson’s disease by combining
serial analysis of gene expression and genetic linkage. Hum
Mol Genet 2003;12:671–677.
78. Noureddine MA, Li YJ, van der Walt JM, et al. Genomic convergence
to identify candidate genes for Parkinson disease: SAGE
analysis of the substantia nigra. Mov Disord 2005;20:1299–1309.
79. Lesnick TG, Papapetropoulos S, Mash DC, et al. A genomic
pathway approach to a complex disease: axon guidance and Parkinson
disease. PLoS Genet 2007;3:e98.
80. Lesnick TG, Sorenson EJ, Ahlskog JE, et al. Beyond Parkinson
disease: amyotrophic lateral sclerosis and the axon guidance
pathway. PLoS ONE 2008;3:e1449.
81. Khalili K, White MK, Lublin F, et al. Reactivation of JC virus
and development of PML in patients with multiple sclerosis.
Neurology 2007;68:985–990.
82. Gelinas DS, DaSilva K, Fenili D, et al. Immunotherapy for
Alzheimer’s disease. Proc Natl Acad Sci USA 2004;101(suppl
2):14657–14662.
83. Araujo RP, Liotta LA, Petricoin EF. Proteins, drug targets and
the mechanisms they control: the simple truth about complex
networks. Nat Rev Drug Discov 2007;6:871–880.
84. Butcher EC. Can cell systems biology rescue drug discovery?
Nat Rev Drug Discov 2005;4:461–467.
85. Palacios R, Goni J, Martinez-Forero I, et al. A network analysis
of the human T-cell activation gene network identifies
JAGGED1 as a therapeutic target for autoimmune diseases.
PLoS ONE 2007;2:e1222.
86. Fernald GH, Knott S, Pachner A, et al. Genome-wide network
analysis reveals the global properties of IFN-beta immediate transcriptional
effects in humans. J Immunol 2007;178:5076–5085.
87. Materi W, Wishart DS. Computational systems biology in drug
discovery and development: methods and applications. Drug
Discov Today 2007;12:295–303.
88. Aradi I, Erdi P. Computational neuropharmacology: dynamical
approaches in drug discovery. Trends Pharmacol Sci 2006;27:
240–243.
89. Baranzini SE, Bernard CC, Oksenberg JR. Modular transcriptional
activity characterizes the initiation and progression of autoimmune
encephalomyelitis. J Immunol 2005;174:7412–7422.
138 Annals of Neurology Vol 65 No 2 February 2009
90. Woolfe F, Waxman SG, Hains BC. In silico modeling of axonal
reconnection within a discrete fiber tract after spinal cord
injury. J Neurotrauma 2007;24:421–432.
91. Yildirim MA, Goh KI, Cusick ME, et al. Drug-target network.
Nat Biotechnol 2007;25:1119–1126.
92. Danna EA, Nolan GP. Transcending the biomarker mindset:
deciphering disease mechanisms at the single cell level. Curr
Opin Chem Biol 2006;10:20–27.
93. Nicholson JK. Global systems biology, personalized medicine
and molecular epidemiology. Mol Syst Biol 2006;2:52.
94. Ergun A, Lawrence CA, Kohanski MA, et al. A network biology
approach to prostate cancer. Mol Syst Biol 2007;3:82.
95. Ray S, Britschgi M, Herbert C, et al. Classification and prediction
of clinical Alzheimer’s diagnosis based on plasma signaling
proteins. Nat Med 2007;13:1359–1362.
96. Corvol JC, Pelletier D, Henry RG, et al. Abrogation of T cell
quiescence characterizes patients at high risk for multiple sclerosis
after the initial neurological event. Proc Natl Acad Sci
USA 2008;105:11839–11844.
97. Fogel LJ. Biotechnology: concepts and applications. Englewood
Cliffs, NJ: Prentice-Hall, 1963.
Villoslada et al: Systems Biology in Neurology 139