Application of systems modelling in drug discovery; examples of impact and future direction

Neil Benson
(Pfizer UK)

Abstract:
Background: reviews of the productivity of the pharmaceutical industry have concluded that the current business model is not optimal, or even unsustainable [2]. Various remedies for this have been proposed, however, these do not directly address the fundamental issue; namely that it is the knowledge required to enable good decisions in the process of delivering a drug that is largely absent; in turn, this leads to a disconnect between our intuition of what the right drug target is and the reality of pharmacological intervention in a system such as a human disease state. As this system is highly complex, modelling will be required to elucidate emergent properties together with the data necessary to construct such models. Currently, however, both the models and quality data for constructing such models are sparse.
Ob jective: the ultimate solution to the problem of pharmaceutical productivity may be the virtual human [1]. However, it is likely to be many years before this goal is realised in full. The current challenge is therefore whether systems modelling can contribute to improving productivity in the pharmaceutical industry in the interim and help to guide the optimal route to the virtual human.
Methods: the examples discussed in this presentation will use ordinary differential equation (ODE) models to draw conclusions about optimal drug targets in given pathways, the biomarker responses that could be used to quantify a successful outcome, the impact of pharmacology and pharmacokinetics and the influence of access to a target. Sensitivity analysis and experiments to test model assumptions have been used to understand and validate models. Results; models have been applied in the pain, anti-infective and allergy disease areas, in particular relating to identifying optimal drug modalities, understanding drug penetration to site of action in the brain and identifying novel targets in the hepatitis C virus/host interaction.
Conclusions: systems modelling has contributed to improved decision making in projects, by identifying approaches carrying a high risk of failure, highlighting optimal biomarkers of outcome and generating hypotheses that can be tested early in drug discovery. A key component for success was the formation of clear questions at the outset of modelling. Future directions and the need for creating a dynamic interaction between model building and data generation will be discussed.

[1] P. Kohl and D. Noble. Systems biology and the virtual physiological human. Molecular and Systems Biology, 5(292), 2009.
[2] B. Munos. Lessons from 60 years of pharmaceutical innovation. Nature Reviews: Drug Discovery, 8(12):959-968, 2009.

Back

Unraveling The Spatial Regulation of Cd95-Induced Apoptosis

Roland Eils
(DKFZ Heidelberg)

Abstract:
Apoptosis occurs through a tightly regulated cascade of caspase activation. Upon induction of death receptors, the most upstream initiator caspase-8 is activated and released into the cytosol. Different models have been proposed for the function of this release for its activity. I will present a novel approach based on localization probes that allows to quantitatively characterize spatial-temporal activity of caspases in living single cells. Our study reveals that while most caspase-8 substrates are cleaved with a similar efficiency, the activity leading to caspase-8 cytosolic release is much lower. Caspase-8 is mostly active at the plasma membrane and targeting of caspase-8 substrates to the plasma membrane can significantly accelerate cell death. In contrast, probes that contain caspase-8 prodomain cleavage sequence targeted to the plasma membrane are not cleaved, indicating a local reaction. Thus, compartmentalization of enzymes and substrates is used by cells to fine tune apoptotic response and offers new possibilities to interfere with cellular apoptotic sensitivity.

Back

Toward a molecular, cellular and systems-based theory of the adaptive immune response

Philip Hodgkin
(Walter Eliza Hall Institute)

Abstract:
When lymphocytes, the primary mediators of immunity, are stimulated they proliferate and the receipt of soluble and cell contact mediated signals help regulate their rate of growth, survival and differentiation. This complex system is well suited to experimental dissection, and offers a useful testing ground for developing concepts in systems biology.
Quantitative experiments tracking individual cell fates reveal relatively simple cellular rules operating independently within each cell. We have used time-lapse video microscopy of stimulated lymphocytes to examine times to divide and die, and more recently times to differentiate of similar cells placed under identical conditions [1]. These experiments reveal an extraordinary degree of variability in the behaviour of individual cells that, irrespective, consistently leads to a high degree of predictability in the population outcomes [2], [3]).
These and other data are consistent with a theory whereby the adaptive immune response is effectively programmed by manipulating a small number of regulable, stochastic internal cellular mechanisms, that when deployed in different ways are able to govern and tune the emergence of a broad range of heterogeneous cell fates. Furthermore, it seems reasonable to speculate the stochastic parameters associated with cell fate decisions have been optimized through natural selection to best advantage the vertebrate immune response to successfully respond to multiple different categories of pathogen.

[1] E. D. Hawkins, J. F. Markham, L. P. McGuinness, and P. D. Hodgkin. A single-cell pedigree analysis of alternative stochastic lymphocyte fates. Proceedings of the National Academy of Sciences, 106(32):13457-13462, 2009.
[2] E. D. Hawkins, M. L. Turner, M. R. Dowling, C. van Gend, and P. D. Hodgkin. A model of immune regulation as a consequence of randomized lymphocyte division and death times. Proceedings of the National Academy of Sciences, 104(12):5032-5037, 2007.
[3] V. G. Subramanian, K. R. Duffy, M. L. Turner, and P. D. Hodgkin. Determining the expected variability of immune responses using the cyton model. Journal of Mathematical Biology, 56(6):861-892, 2008.

Back

Listening to the noise: stochastic fluctuations in molecular biology

Mustafa Khammash
(University of California Santa Barbara)

Abstract:
One of the challenges in the analysis and synthesis of genetic networks is that the cellular environment in which these circuits function is abuzz with noise. Cellular noise results in random fluctuations (over time) within individual cells and is a potential source of phenotypic variability among clonal cellular populations. Cells appear to exploit their noisy environment in some instances and to suppress its effects in others. Whatever the role of noise, the underlying cellular architecture cannot be fully understood without taking stochastic noise effects into account. In this talk, we present new tools for the analysis of noisy genetic networks and demonstrate their application for specific biological examples. We show that noise induced fluctuations carry within them valuable information about network architecture. Far from being a nuisance, the ever-present cellular noise acts as a rich source of excitation that propagates through a given network and has its statistics shaped by that network's distinctive architecture. When measured using single cell techniques, fluctuation statistics can be processed to yield information about network connectivity and parameters that cannot be obtained using deterministic identification means.

Back

Pharmacometrics and systems biology: do they fit?

Charlotte Kloft
(Martin-Luther-University Halle-Wittenberg)

Abstract:
In drug development and therapeutic use of drugs the two aspects of increasing costs and pressure of time are currently faced. Especially oncology presents a field with complex and individual disease status and progression as well as with a limited number of anticancer drugs being administered in often empirical dosing regimens and leading to highly variable success rates. Hence, understanding of the patients' status and the disease ('system') and the development of new drugs and definition of tailored treatment regimens seem appealing for more rationale and streamlined approaches.
Cancer is mainly caused by alterations in oncogenes and/or tumour-suppressor genes resulting in changes in growth regulation. Technological advances especially on a molecular level in the last decade have gained a better understanding of growth factors and the interaction with their receptors and the down-stream signal transduction on the transformation of healthy/normal cells into maligne cells and on regulation of cell proliferation and/or apoptosis. The knowledge in these areas has improved the abilities to diagnose and classify cancer entities.
The fundamental challenge of anticancer drug treatment is to effectively eliminate cancer cells by a dosing regimen that is being tolerated by the patient. Most classical cytotoxic anticancer agent do not distinguish between healthy and cancer cells and thus inhibit essential functions in both cell populations. A new generation of anticancer drugs has been designed to interfere with specific molecular targets that have been identified a critical role in tumour growth or progression. In recent years, a variety of new generation drugs has been developed targeting specific structures on the cell membranes (e.g. monoclonal antibodies) or oncogenic proteins in cancer cells (e.g. small molecule tyrosine kinase inhibitors). Pharmacometrics aims to elucidate the relationships between dosing-drug exposure-resulting drug effects (killing of cancer cells/toxic site effects) while incorporating knowledge of the patients and their disease.
The presentation will (i) address special features of drug cancer treatment and the current approaches of data analysis in oncology, (ii) illustrate in case example the application of emerging tools in modelling and simulation, (iii) highlight the improvements of cross-disciplinary interaction and (iv) finally indicate future challenges to be met.
In future, effective integration of knowledge in biomarkers for disease, pharmacokinetics and pharmacodynamics, as well as combining approaches in systems biology and modelling/simulation might streamline drug development and increase the risk/benefit ratio of therapeutic use in the individual patient.

[1] C. Kloft, J. Wallin, A. Henningsson, E. Chatelut, and M. O. Karlsson. Population Pharmacokinetic-Pharmacodynamic Model for Neutropenia with Patient Subgroup Identification: Comparison across Anticancer Drugs . Clinical Cancer Research, 12(18):5481-5490, 2006.
[2] B.-F. Krippendorff, K. Kuester, C. Kloft, and W. Huisinga. Nonlinear Pharmacokinetics of Therapeutic Proteins Resulting From Receptor Mediated Endocytosis. Journal of Pharmacokinetics and Pharmacodynamics, 36(3):239-260, 2009.

Back

Function by design: how emergent design properties shape the biological function of the RAS-RAF-MEK-ERK/MAPK

Walter Kolch
(Systems Biology Ireland)

Abstract:
The RAS-RAF-MEK-ERK pathway controls a variety of fundamental cellular processes including cell proliferation, transformation, differentiation, motility and apoptosis. While individual biochemical steps are vey well studied, it still remains enigmatic how this pathway can mediate so many diverse biological functions with high specificity and fidelity. Here, we will discuss how design properties of the ERK pathway determine its regulation and biological function analysing the following three aspects:
1) Analysing pathway topology by Bayesian inference. Topology can have a decisive influence on the activation kinetics and biological function of a pathway. Although we can describe complex networks in terms of components, we cannot easily determine which topology the cell actually uses to transduce a specific signal. Using Bayesian inference-based modelling we can use measurements of a limited number of biochemical species combined with multiple perturbations to reliably distinguish between different plausible pathway topologies.
2) The role of feedback loops in shaping ERK response kinetics. The three-tiered RAF-MEK-ERK kinase cascade functions as an amplifier. However, this pathway also contains a number of negative feedback loops that in combination with the amplifier module produce interesting effects, for instance increasing drug resistance, constraining responses within a cell population, and conveying robustness against perturbations in the amplifier.
3) The role of noise. The ERK pathway converts receptor signals that can be of variable and fluctuating strengths into robust biological responses. Thus, there must be mechanisms to filter true signals from noise. We have analysed such mechanisms by a combination of biochemical measurements and mathematical (deterministic and probabilistic) modelling. We find that the activation of ERK is different between individual cells and cell populations depending on dose and duration of the input. Explicitly incorporating protein expression variability in kinetic models, we demonstrate that analogue behaviour on the single cell levels can generate digital output on the population level. This conversion relies on stochastic noise in protein expression between individual cells and the existence of a negative feedback loop from ERK to Ras/Raf-1 that makes responses of individual cells more graded. Thus, systems responses, which depend on a complex interplay between pathway topologies, the interaction kinetics, and activation dosage are dramatically shaped by noise in protein expression.

Back

Assessing key players in disease relevant pathways

Ursula Kummer
(University of Heidelberg)

Abstract:
Modeling the dynamics of biochemical systems has become as essential and integral part of systems biology. The rational behind this fact is that a true understanding of biochemical networks must eventually result in valid predictions about the behaviour of these network in the living cell. In most cases, such predictions will only be possible in the context of concise descriptions of the networks which allow a thorough analysis. Mathematical descriptions are doubtless concise and allow for a multitude of ways to analyse their behaviour. Thus, the identification of key players by conducting sensitivity analyses has been suggested and used for drug target identifications in the past.
However, in general, a lot of precise experimental data is needed to build a quantitative, computational models with predictive power. Still, sensitivity analyses are often conducted locally, for one specific set of parameters. This poses a problem for the application in drug target identification. Thus, not knowing many of the parameters asks for more global approaches.
In this presentation, I will introduce different strategies for a global assessment of key players in biochemical pathways. As an example a systems biology investigation of interferon signalling is presented. This study identified IRF-9 as a key player in the signalling pathway which is crucial in anti-viral defense.

Back

Multi-Pathway Network Analysis of Epithelial Cell Responses in Inflammatory Environments

Douglas Lauffenburger
(Massachusetts Institute of Technology)

Abstract:
Cellular responses to external stimuli depend upon dynamic features of multi-pathway network signaling, and thus ultimate phenotypic behavior is influenced in a complex manner by both cellular environment and intrinsic cellular properties. We have established across the past half-decade that multi-variate systems analysis employing 'data-driven' computational modeling methods aimed at quantitative experimental data within a 'cue-signal-response' paradigm are capable of comprehending these convoluted effects for relatively simplified in vitro cell culture examples. In this Workshop contribution, I will present recent extensions of this foundational work along avenues including emphasis on application to in vivo contexts and on connection to interactome databases, focused on epithelial cell responses in inflammatory environments.

Back

Regulatory T cell control of pathogenic and effector T cells - implications for the development of new therapeutics for cancer and autoimmunity

Kingston Mills
(Trinity College Dublin)

Abstract:
Immune homeostasis is dependent on a balance between regulatory and effector/pathogenic T cells and manipulating this balance may be a promising approach for the treatment of cancer and autoimmunity. It is now well established that IL-17-producing CD4+ T cells (Th17 cells) play a pathogenic role in many autoimmune diseases, a function previously attributed exclusively to IFN-γ -secreting Th1 cells. Furthermore, Th17 cells can function with Th1 cells to mediate protective immunity to infection and may also be important in immunity to tumors. Regulatory T (Treg) cells can suppress pathogenic T cells directed against self-antigens and thereby prevent autoimmunity, and limit immunopathology during infection, but can also suppress protective immunity to pathogens and to tumors. Stimulation of innate immune cells, especially dendritic cells (DC), by pathogen-associated molecular patterns (PAMPs) through TLRs, NLRs, RLRs and CLRs, provide the signals for maturation and cytokine production (signals 2 and 3) for T cell activation. However other innate cells, including NK cells and γδ T cells, also play a role in induction of Th1 and Th17 cells respectively. We have shown that TLR and NLR agonists induce IL-1 α and IL-1 β as well as IL-23, which synergize to promote activation of memory Th17 cells. IL-1 and IL-23 also stimulate IL-17 production by γ δ T cell, which act in an autocrine manner to stimulate activation of Th17 cells. The induction of function of Th17 cells is regulated by cytokines secreted by the other ma jor subtypes of T cells, especially IL-10 and TGF-β production by Treg cells. The induction of natural and adaptive Treg cells is stimulated by TGF-β , IL-10 and IL-27 in response to certain virulence factors from pathogens that have evolved to subvert protective immunity, but also in response to TLR and NLR agonists and this constrains their potential as adjuvants and therapeutics against infection and cancers. We have examined the signalling pathways in PAMP-activated DC that promote the innate cytokines that drive T cell responses. We have shown that inhibitors of MAP kinase and PI3K signalling pathways can block IL-10 and TGF-β , while sparing or enhancing IL-12 production, thereby promote Th1, but not Treg cells. This has been successfully applied to a therapeutic intervention against tumors in pre-clinical models. finally, we have identified approaches for activation of regulatory cytokine production by innate immune cells for selective induction of Treg cells, without Th1 or Th17 cells. These approaches been effective in pre-clinical models of autoimmunity and cancer and are also being translated to clinical studies in humans.

Back

Multiscale Modeling of Hepatitis C Virus Infection

Alan Perelson
(Los Alamos National Laboratory)

Abstract:
Mathematical models of viral infection and treatment for illnesses caused by agents such as HIV, hepatitis C virus (HCV) and hepatitis B virus, have been successful in uncovering basic features underlying the pathogenesis of these diseases, such the rate of viral production in infected individuals and the lifespan of infected cells. In the case of HCV infection, new direct acting antiviral agents that interfere with various intracellular molecular processes are in clinical trials. In order to model the action of these agents, new multiscale models that simultaneously consider intracellular events as well events at the cellular and extracellular level are needed. I will discuss our recent progress in this area and show that certain features of HCV infection that we had previously deduced based on analyzing patient responses to interferon and ribavirin therapy may have to be revised based on new data and these new multiscale models. Further, drug resistance to HCV protease inhibitors when used as monotherapy has appeared in clinical trials more rapidly than ever seen before in the treatment of any infectious disease - as early as the second day of therapy 5-20% of viral isolates have been shown to be resistant. Models have revealed how this is possible but have also raised some unanswered questions about the rapid in vivo growth of drug resistant virus.

Back

Long-Lasting Therapeutic Effects of Desynchronizing Brain Stimulation

Peter Tass
(Institute of Neuroscience and Medicine, Jülich)

Abstract:
A number of brain diseases, e.g. movement disorders such as Parkinson's disease, are characterized by abnormal neuronal synchronization. Within the last years permanent high-frequency (HF) deep brain stimulation became the standard therapy for medically refractory movement disorders. To overcome limitations of standard HF deep brain stimulation, we use a model based approach. To this end, we make mathematical models of affected neuronal target populations and use methods from statistical physics and nonlinear dynamics to develop mild and efficient control techniques. Along the lines of a top-down approach we test our control techniques in oscillator networks as well as neural networks. In particular, we specifically utilize dynamical self-organization principles and plasticity rules. In this way, coordinated reset (CR) stimulation [1], an effectively desynchronizing brain stimulation technique, has been developed. The goal is not only to counteract pathological synchronization on a fast time scale, but also to unlearn pathological synchrony by therapeutically reshaping neural networks [2]. I shall present the theory, results from both experiments in MPTP monkeys and a first clinical application. CR stimulation can also be applied non-invasively, e.g. as acoustic CR stimulation for the treatment of tinnitus. The latter is characterized by abnormal delta band synchronization e.g. in the auditory cortex. I shall present results from a single-blinded multicenter clinical trial in 63 patients.

[1] P. Tass. A model of desynchronization deep brain stimulation with a demand-controlled coordinated reset of neural subpopulations. Biological Cybernetics, 89:81-88, 2003.
[2] P. Tass and M. Majtanik. Long-term anti-kindling effects of desynchronizing brain stimulation: a theoretical study. Biological Cybernetics, 94:58-66, 2006.

Back