Abstract:  Since many years, computers have played an important role in helping scientists to store, manipulate, and analyze data coming from many different disciplines. In recent years, however, new technological capabilities and new ways of thinking about the usefulness of computer science is extending the reach of computers from simple analysis of collected data to hypothesis generation. The aim of this work is to provide a contribution in the Computational Systems Biology field. The main purpose of this recent discipline is to enhance the intertwined relationship connecting Biology and Computer Science, by developing tools and theoretical frameworks able to formally and quantitatively investigate the interactions among the components of biological systems. The final goal of these efforts is to assemble the different pieces into a working model of a living, responding, reproducing cell; a model that can be used for performing in-silico tests and simulations in order to understand and predict possible emergent properties. In this thesis we present the application on real biological case studies of a specific concurrent modelling language (derived by the metaphors of "molecules-as-objects" -introduced by Fontana- and "cells-as-computations" -introduced by Regev and Shapiro- at the end of last century) and the development and implementation of a tool for inferring knowledge from experimental data in order to link the numerical aspects of a model to real wet-lab data.

Abstract: Biology is the science of life and living organisms. This discipline has seen a tremendous growth during the last decades, powered by the deployment of automated experimental frameworks. Recently, increasing attention has been put on the convergence between Biology and diverse disciplines, among which Computer Science. Because of its potentialities to disentangle major biological issues at the system level, computational thinking is gaining more and more relevance. Moreover, these research areas have started to inform each other and researches began transgressing disciplinary boundaries.
In this thesis we presented methods and approaches to tackle the problem of knowledge discovery in Computational Biology from a stochastic perspective. Major bottlenecks in adopting a stochastic representation can be overcome with the use of proper methodologies by integrating statistics and computer science. In particular we focused on parameter inference for stochastic models and efficient model analysis. We showed the application of these approaches on real biological case studies aiming at inferring new knowledge even when a priori (and/or experimental) information is limited.

Abstract: The identification of genes acting synergistically as master regulators of physiologic and pathologic cellular phenotypes is a key open problem in systems biology, Here we use a computational reverse-engineering approach to identify the repertoire of transcription factors (TFs) of a master regulatory module responsible for synergistic activation of a tumor-specific signature. Specifically, we used the ARACNe algorithm and other computational tools to infer regulatory interactions responsible for initiating and maintaining the mesenchymal phenotype of Glioblastoma Multiforme (GBM), previously associated with the poorest disease prognosis. Expression of mesenchymal genes is a hallmark of aggressiveness but the upstream regulators of the signature are unknown. Starting from the unbiased analysis of all TFs, we identify a highly interconnected module of six TFs jointly regulating >75% of the genes in the signature. Two TFs (Stat3 and C/EBPb), in particular, display features of initiators and master regulators of module activity. Biochemical validation confirms that the TFs in the module bind to the inferred promoters in vivo and ectopic expression of the master TFs activates expression of the mesenchymal signature. These effects are sufficient to trigger mesenchymal transformation of neural stem cells, which become highly tumorigenic in vivo, and promote migration and invasion. Conversely, silencing of Stat3 and C/EBPb in human glioma cells leads to collapse of the mesenchymal signature and reduction of tumor aggressiveness. Our results reveal that activation of a small transcriptional module is necessary and sufficient to induce a mesenchymal phenotype in malignant brain tumors. Interestingly, interrogation of pathways upstream of C/EBP and Stat3 identified two new genetic alterations in high-grade glioma patient that jointly explain 54% of the tumor subtype. Such a Pathway Wide Association Study (PWAS) is a critical departure from more traditional Genome Wide Association Studies ( GWAS), which suffer from dramatic power loss due to multiple hypothesis testing.

Abstract: 

Mechanistic, computational models are specifically constructed to illuminate the interpretation of the data upon which they are built. A computational model represents a modeler's understanding of the structure and function of part of a biological system. As the number of researchers constructing quantitative models continues to grow, and as the models become ever more sophisticated, they constitute a significant accumulation of knowledge about the structural and functional organization of biological systems. Moreover, the assimilation of new hypotheses and data into existing models can be done in a more systematic fashion because the additions must fit into the existing constructs using the same rules as for the models themselves. Computational models can thus be far more useful than just encapsulating one modeler's abstraction of a particular system: once properly constructed, the models become a dynamic representation of our current state of understanding of a system in a form that can facilitate communication between research groups and help to direct further investigations. A limitation faced in trying to reuse models is the degree of incompatibility between software tools.  Most software systems have their own unique, specialized representation formats that are tailored to their particular needs.  Likewise, many efforts to develop advanced biological modeling languages result in powerful and expressive formats that usually are implemented only in a single software environment.  Thus, still we face the problem of exchanging and archiving models in a common representation format. SBML has emerged as a candidate lingua franca for communicating many essential aspects of many classes of models in computational systems biology. The Systems Biology Markup Language (SBML) format is a machine-readable model representation language for software tools in computational systems biology; it is the most well-established format for models in the field, with the support of over 200 software tools worldwide today.  However, despite its utility and broad acceptance, there remain numerous aspects of model representation and use that are outside of its current scope. In this presentation, I will discuss ongoing efforts to standardize additional areas of model representation in order to make models and simulation results as reproducible as possible.  Other efforts include MIRIAM (Minimum Information Requested in the Annotation of Models), a set of guidelines for model curation; SBO (Systems Biology Ontology), an ontology of mathematical concepts and rate laws used in computational models; and SED-ML (Simulation Experiment Description Markup Language), a format for expressing simulation experiments in a simulator-independent form, so that different systems can reproduce the steps needed to take a model from initial conditions to a plotted result.  I will also touch on relationships to other standardization efforts.

 
Abstract The reason why Process Algebra has been so successful is because it formulates abstract mathematical equations that describe general properties of an important class of phenomena in the real world. The equations enable us to explain, reason, predict the phenomena, and control them to serve our needs and interests. Originally, Process Algebra has been applied primarily to what happens inside computers and computer networks. Now the algebra has been extended and applied to biological phenomena, at various scales of granularity and abstraction. I will illustrate these points by drawing on the history of CCS, CSP, the pi calculus, Kleene Algebra and Separation Logic, and speculate about the unification of these algebras for future development and application.

Abstract Nucleic acids (DNA/RNA) encode information digitally, and are currently the only truly ‘user-programmable’ entities at the molecular scale. They can be used to manufacture nano-scale structures, to produce physical forces, to act as sensors and actuators, and to do computation in between. Eventually we will be able to use them to design nanostructure at the bottom end of Moore’s Law, and to interface them with biological machinery to detect and cure diseases at the cellular level under program control. The basic technology to create and manipulate these devices has existed for many years, but the imagination necessary to exploit them has been evolving slowly. Recently, some very simple computational schemes have been developed that are autonomous (run on their own once started) and involve only short (easily synthesizable) DNA strands with no other complex molecules. We now need programming abstractions and tools that are suitable for molecular programming, and this requires a whole hierarchy of concepts to come together. Low-level molecular design is required to produce molecules that interact in the desired controllable ways. On that basis, we can then design various kinds of ‘logic gates’ and ‘computational architectures’, where much imagination is currently needed. We also need programming languages to organize complex designs both at the level of gate design, and at the level of circuit design. Since DNA computation is massively concurrent, some tricky and yet familiar programming issues arise: the need to formally verify circuit designs to avoid subtle deadlocks and race conditions, and the need to design high-level languages that exploit concurrency and stochasticity.

Abstract The first step towards a more individualized anti-cancer pharmacotherapy was taken decades ago with the introduction of the estrogen receptor assay and the anti-estrogen tamoxifen for the treatment of breast cancer. The development of tamoxifen was based on knowledge of the pathophysiology of the disease and the mechanism of the action of the drug. Since then we have only seen a few combinations of predictive biomarker tests and targeted anticancer drugs being introduced. The most successful of these combinations is the immunohistochemical assay for the Human Epidermal growth factor Receptor 2 (HER2) and the monoclonal antibody trastuzumab used for treatment of HER2 overexpressing breast and gastric cancer. The relatively recent development of the monoclonal antibodies cetuximab and panitumumab, directed toward the Epidermal Growth Factor Receptor (EGFR), for the treatment of colorectal cancer has taught us that detection of the target for the anticancer drug is not always sufficient to predict the response in tumors with multiple molecular alterations. It was subsequently shown that only patients with non-mutated K-ras in the downstream signaling pathways responded to this treatment. Development of effective anticancer drugs requires an in-depth understanding of the mechanism of action as well as of the pathophysiology with which the drug interacts in the individual patient. Improvement of our understanding of the pathways and network linked to the drug target and to the response loops induced by the pharmacological intervention is crucial in order to succeed in anticancer drug development.

Abstract: The challenge for biology and medicine in the 21st century is the need to deal with its incredible complexity. One powerful way to think of biology is to view it as an informational science requiring systems approaches. This view leads to the conclusion that biological information is captured, mined, integrated by biological networks and finally passed off to molecular machines for execution. Systems approaches are holistic rather than atomistic—and employ both hypothesis-driven as well as discovery-driven approaches. Hence the challenge in understanding biological complexity is that of using systems approaches to deciphering the operation of dynamic biological networks and molecular machines across three time scales of life—development, physiological and disease responses. I will focus on our efforts at a systems approach to disease—looking at prion disease in mice. We have recently published a study that has taken more than 5 years—that lays out the principles of a systems approach to disease including dealing with the striking signal to noise problems of high throughput biological measurements and biology itself (e.g. polymorphisms). I will also discuss the emerging technologies that will transform medicine over the next 10 years—including next generation DNA sequencing, microfluidic protein chips, single-cell analyses and the capacity to generate stem cells for each individual patient. Finally, I will talk about the computational challenges that we are facing with these explosions of data—signal to noise, multiscale integration, visualization of complexity, how to analyze millions of complete genome sequences with their attendant phenotypic data, challenges arising from the human proteome project, etc. It appears that systems medicine, together with pioneering changes such as next-generation DNA sequencing, blood protein measurements (nanotechnology) and stem cell biology, as well as the development of powerful new computational and mathematical tools will transform medicine over the next 5-20 years from its currently reactive state to a mode that is proactive (P4)—medicine that is predictive, personalized, preventive and participatory. P4 medicine will have striking implications for healthcare costs as well as leading to a transformation of the healthcare industry.

Abstract: Molecular interactions are wired in a fascinating way resulting in complex behavior of biological systems. Theoretical modeling provides a useful framework for understanding the dynamics and the function of such networks. The complexity of the biological networks calls for conceptual tools that manage the combinatorial explosion of the set of possible interactions. A suitable conceptual tool to attack complexity is compositionality, already successfully used in the process algebra field to model computer systems. We rely on the BlenX programming language, originated by the beta-binders process calculus, to specify and simulate high-level descriptions of biological circuits. A biological example on regulation of the circadian clock is presented as a case study of compositional modeling.

Abstract: The talk will describe the VENUS-C project and how it fits in the European Cloud Computing landscape. Cloud computing is a newcomer in the European distributed computing infrastructure for science, where dominant technologies so far have been supercomputing, cluster and grid computing. The talk will try to compare the experience of supporting scientific communities and computational system biology in particular with the different computing paradigms. The contribution of CoSBi to VENUS-C will also be presented.

Abstract: The convergence between computer science and biology occurred in successive waves involving deeper and deeper concepts of computing. The current status makes computer science a suitable candidate to become a foundation for systems biology with the same importance as mathematics, chemistry and physics. However, this great opportunity is not a free lunch. New development and a strong integration of different fields of computing are needed to cope with the challenges of systems biology: a complex and expanding applicative domain that can open completely new avenues of research in computing and eventually help it become a natural, quantitative science.

Abstract: Cells use a dense network of signaling pathways to decide how to respond to various external stimuli. Several dynamic aspects of complex pathways have been already described. Here we show that simple generic motifs of signaling pathways (without any feedback) could show some interesting dynamics. We investigated the dynamics of the simplest dynamical elements in biochemical networks: we analyzed the response dynamics of a signaling protein when it enters the signaling pool in one state (modified or unmodified) and exits in both of these states. When the exit rates of these two states are comparable, a persistent stimulus results in step responses and can produce ultrasensitivity, however, when the exit rates are imbalanced, the signaling protein gives transient responses to persistent stimuli. Such adaptive behavior of signaling pathways could be used by many organisms. We also investigated the dynamical features of phosphorelays: phosphorelays are extended two-component signaling systems found in diverse bacteria, lower eukaryotes and plants. We found that the intermediate layers of phosphorelays can display ultrasensitivity that could result in tolerance of pathway cross-talk. Furthermore, it leads to a high signal to noise ratio for the relay output. We show that these features of phosporelays might be employed by the sporulation network of B.subtilis

Abstract: A common practice at CoSBi, and in many other research centers working in systems biology, is to test new paradigms, ideas and algorithms by building specific software tools tailored to solve the problem at hand. Over the last year, CoSBi has begun to bring these different prototypes together under one consistent software framework: CosbiLab. At high level, CosbiLab is a "Systems Biology Studio": a collection of  applications, libraries, components which purpose is to help scientists in their research in systems biology. More precisely, the goal of CosbiLab is to support the scientific method underlying systems biology, in particular focusing on the in-silico study of the dynamics of biological systems.
In the simplest scenario, a Systems Biology researcher wants to perform a basic in-silico experiment: write a model, simulate it, and visualize the simulation results. They might have an idea of how chemicals interact in a particular pathway, so they want to see if the interactions they have in mind produce the correct behaviour; or they might have a clear idea of the pathway, but do not know what the foreseeable result is when interacting with an additional entity and so on.
CosbiLab v1 will support this basic scenario, providing an integrated way to (a) write a model (Input), (b) simulate it (Processing), and (c) view if the simulated model behaves as expected and perform further analysis on the results (Output).This demo will show an early version of CosbiLab v1.

Abstract: Complex networks of interacting proteins control the physiological properties of a cell (metabolism, reproduction, motility, signaling, etc.). Intuitive reasoning about these networks is often sufficient to guide the next experiment, and a cartoon drawing of a network can be useful in codifying the results of hundreds of observations. But what tools are available for understanding the rich dynamical repertoire of such control systems? Why does a control system behave the way it does? What other behaviors are possible? How do these behaviors depend on the genetic and biochemical parameters of the system (gene dosage, enzymatic rate constants, equilibrium binding constants, etc)? We use the basic principles of biochemical kinetics to convert a network diagram into mathematical rate laws. These rate laws can be used to predict the average properties of the control system by analysis and simulation of nonlinear ordinary differential equations or to predict the particular behavior of individual cells by stochastic simulations (using Gillespie’s algorithm). We have also explored a number of hybrid strategies, combining deterministic and stochastic simulations with discrete and continuous variables. We will illustrate these ideas and strategies by exploring the molecular network controlling the cell division cycle of eukaryotes.

Abstract My vision for the 21st Century: Computational thinking will be a fundamental skill used by everyone in the world. To reading, writing, and arithmetic, we should add computational thinking to every child's analystical ability. Computational thinking involves solving problems, designing systems, and understanding human behavior by drawing on the concepts fundamental to computer science. Thinking like a computer scientist means more than being able to program a computer. It requires the ability to abstract and thus to think at multiple levels of abstraction. In this talk I will give many examples of computational thinking, argue that it has already influenced other disciplines, and promote the idea that teaching computational thinking can not only inspire future generations to enter the field of computer science but benefit people in all fields.

Abstract Current human nutritional and genetic epidemiological methods yield “risk factors” called population attributable risks (PAR). These risk factors are statistical estimates of the percentage reduction in disease in the population if the risk were to avoided or the gene variant were not present – these measures are often assumed to apply to individuals who are likely to differ in genetic make-up, lifestyle, and dietary patterns than those individuals in the study population. Developing individual risk factors in light of the genetic diversity of human populations, the complexity of foods, culture and lifestyle, and the variety of metabolic processes that lead to health or disease are significant challenges for personalizing dietary advice for healthy or medical treatments for individuals with chronic disease. New research and application strategies are needed for creating knowledge for personalizing nutrition advice and healthcare.

Abstract: For more than one century, the measure of drug structure-activity relationships has been based on mathematical equations describing the interaction of the drug with its biological receptor. The understanding of the multiplicity of biological responses induced by the drug-receptor interaction demonstrated the limits of current approach and the necessity to develop novel concepts for the quantitative analysis of drug action.
Current availability of technologies for the creation of animal models where a specific reporter systems enables the rapid and quantitative study of drug action in living animals, opens the way to novel concepts to be applied to drug development. We used a mouse model engineered to measure estrogen receptor (ER) transcriptional activity in living organisms, to investigate the effect of long term (21 days) hormone replacement on ER signaling by whole body, in vivo, imaging. Estrogens and SERMs were administered daily at doses equivalent to those used in humans as calculated by the allometric approach. As controls, ER activity was measured also in cycling and ovariectomized mice. The study demonstrated that ER-dependent transcriptional activity oscillated in time and the frequency and amplitude of the transcription pulses was strictly associated with the target tissue and the estrogenic compound administered. Our results indicate that the spatio-temporal activity of Selective Estrogen Receptor Modulators (SERMs) is predictive of their structure demonstrating that the analysis of the effect of estrogenic compounds on a single surrogate marker of ER transcriptional activity is sufficient to classify families of compounds structurally and functionally related.
Thus a systematic study of spatio-temporal effects may be proposed as a measure of drug efficacy for the classification of pharmacologically active compounds. The application of this methodology is expected to simplify the identification of families of molecules functionally correlated and to speed up the process of drug discovery. (Collaborators: Gianpaolo Rando and Paolo Ciana)

Abstract: In part as a recognition of the ‘4th Paradigm’ in scientific discovery, in recent years Microsoft Research has steadily increased its interest in the application of computer science tools and technologies to breakthrough science. This has happened in diverse areas of research ranging from astronomy to oceanography, from molecular biology to ‘big history’ and from sociology to climatology. We have enjoyed some significant successes by improving access to scientific information, for example with Microsoft Research’s Worldwide Telescope, but have also discovered that there are many challenges remaining when one considers the relatively fragmented context within which most science is undertaken. This presentation explores the ways in which Microsoft is enabling changes in the way science is conducted, it will evidence how some existing tools can be used to improve the way in which data and information are discovered/shared/visualized and gives lessons learned from our collaborative research engagements associated with realizing our vision for the 4th Paradigm.

Abstract: Traditional Computer Architecture always had a bottom-up bias: let us design the processor and let the upper layers of firmware, runtime, compiler, programming model figure out how to use it efficiently. This design practice was grounded in the ability to produce high-performance processors with increasing clock frequencies and design complexity. However, the industry then hit the power-density wall, and switched to incorporating more processing cores (at the same or reduced frequency) in the same die. Unfortunately, the bottom-up design bias still lingers on. At the BSC Microsoft Research Centre we argue for a top-down approach: to let software requirements drive computer architecture innovation forward. In the talk, we will give several examples of our hardware research that is guided by programming models or runtime systems needs.

Abstract: Diabetes millitus type 2 is a metabolic disorder characterized by high blood glucose in the context of insulin resistance. The causes of this insulin resistance are not completly specified but there is increasing evidence to think that there is a relation between mitochondrial function and the insullin signalling. In that sense, our general goal is to clarify the interaction between them by means of system biology approaches. Specifically, the objective of this project is to develop a computational framework that will allow the prediction of functional associations and interactions between insulin signalling and mitochondrial function, which will be tested both in mammalian cells, mouse and drosophila. A major challenge we are doing in collaboration with COSBI center is the combination of signalling and metabolic pathway in a unique network combining the different nodes from both pathways. Moreover, using this network we will be able to simulate different complex metabolic and regulatory scenarios for normal conditions and diseases, and identify and predict possible changes of the system when some parts have been altered.

Abstract: A large set of enumerated elements is a data structure commonly used in many application domains; these applications would benefit from a fast implementation of operations like set intersection or union. A vector processor is suitable for such kind of operations when the set is highly populated. A vector of bits can easily represent the data where each bit indicates the presence or the absence of each element, but it is not appropriate when the density of elements in the set is low. Here we present a way to efficiently vectorize these operations when working with sparse sets using two levels of vectors of bits. The instructions needed to manipulate the two levels are also introduced. Additionally, we present preliminary results of the vectorized intersect operation when it is used in a data mining application.

Abstract:

The Microsoft Research-INRIA Joint Centre was founded by INRIA, Microsoft Corporation, and MSR Laboratory Cambridge. The Centre's objective is to pursue fundamental, long-term research in formal methods, software security, and the application of Computer Science research to the Sciences.
The research programme is divided into two tracks.
- A, Software Security and Trustworthy Computing, comprising 3 projects: Mathematical Components, Tools and Methodologies for Formal Speci?cations and for Proofs, and Secure Distributed Computations and their Proofs.
- B, Computational Sciences and Scienti?c Information Interaction, comprising 4 projects: Dynamic Dictionary of Mathematical Functions, ReActivity, Adaptive Combinatorial Search for E-Science, and Image and Video Mining for Science and Humanities.

I will discuss of some recent results.

Abstract: The object of the Mathematical Components project is to demonstrate that formalized mathematical theories can, like modern software, be built out of components. To validate this approach, and the resulting body of formalized mathematical components, we are carrying out the proof of the Odd Order theorem with the Coq system and it's extension ssreflect.
In this talk I'll give an overview of the technology we use and develop, and report on the status of the formalization.