ICSB 2017: INTERNATIONAL CONFERENCE ON SYSTEMS BIOLOGY 2017
PROGRAM FOR FRIDAY, AUGUST 11TH
Days:
previous day
all days

View: session overviewtalk overview

08:30-10:30 Session 20: Friday Morning
Location: Colonial Hall
08:30
COPASI – Application in research and training
SPEAKER: Ursula Kummer

ABSTRACT.  

COPASI is a software for computational systems biology that is and has been developed as a joint venture of the groups of Pedro Mendes (now at University of Connecticut) and my own group since the year 2000. During these years COPASI has integrated standard, as well as newly developed algorithms to model, simulate and analyse biological processes in many different ways. Thus, COPASI supports e.g. both ODE-based as well as stochastic (discrete particle based) formalisms. In addition, COPASI features include steady state calculation, sensitivity analyses, optimization, parameter fitting, parameter scanning and computing features of nonlinear dynamical systems such as Lypaunov exponents, among others. COPASI houses new developments, e.g. in the area of hybrid approaches, parameter estimation, complexity reduction and nonlinear dynamics. In essence, we use COPASI as a platform in which we integrate new algorithms and methods once they are validated and relatively stable to use. 

Due to the fact that the basic features are simple to use, the software is also often used in the training of systems biologists around the world. However, the software is certainly not restricted to basic, simple features and advanced features are used in many different research projects. 

The talk will briefly summarize the features of the software and will then concentrate on recent advanced developments and their application in research. Finally, one example application on the multi-scale modeling of NFkB signalling in hepatocytes is described in more detail.

09:00
Systematic integration of models and data for yeast growth and division
SPEAKER: Edda Klipp

ABSTRACT. With the progress of genome-wide experimental approaches we witness the establishment of more and more libraries of genome-wide data for proteins or RNA or metabolites. However, the separated consideration of metabolic networks or gene regulation networks does not tell us how these networks are integrated to allow a cell to grow, divide and respond to changing environments.

We use the yeast Saccharomyces cerevisiae as the model organism for eukaryotic cells allowing to comprehensively analyzing regulatory networks and their integration with cellular physiology. We use a modular and iterative approach that allows for a systematic integration of cellular functions into a comprehensive model allowing to link processes that are strongly interlinked in cellular life, but measured separately.

 

09:30
Friday Morning Coffee Break
SPEAKER: Coffee Break
10:00
A model-driven experimental approach unveils the interplay between the circadian factor Period 2 and the tumor suppressor p53.

ABSTRACT. In addition to coordinating the body’s physiological and metabolic functions, the circadian (~24h) core clock mechanism exerts a multilevel regulation of the cell division process. This is particularly relevant when it comes to the mechanism by which circadian factors, such as PER2, modulate the cellular response to genotoxic stress that leads to cell cycle arrest. We have previously reported that PER2 directly interacts with the C-terminus domain of p53 promoting p53’s stabilization. Interestingly, PER2 binding to p53 prevents p53 oligomerization and its transcriptional activity in response to genotoxic stress. Unexpectedly, the expression peaks of PER2 and p53 were found out-of-phase in total extracts, a result that was in sharp contrast with the predicted stabilizing role of PER2 on p53. In this talk, we will illustrate how these unexpected phase relationships were analyzed by using systematic modeling of all possible regulatory scenarios to predict the ou-of-phase relationship between PER2 and p53 were possible under conditions in which i) PER2 association to p53 favors p53’s nuclear entry and where ii) PER2 was able to bind to various ubiquitinated forms of p53. Model predictions were validated by findings that showed p53 half-life was increased in in the nucleus compared to the cytosol compartment and its localization modulated by endogenous PER2 levels. Thus, overexpression of PER2 drove p53 to the nucleus whereas PER2 depletion by siRNA prevented p53’s nuclear accumulation. These results were further supported by in vitro binding assays that confirmed PER2 binding to p53 occurred independently of p53’s ubiquitination status. In summary, our work illustrates how clock regulatory nodes can be inferred from oscillating time course data with the combination of mathematical modeling and experimental work.

10:30-12:30 Session 21A: Parallel Session VII a: Systems Biology Education
Chair:
Location: Colonial Hall
10:30
The Undergraduate Degree in Systems Biology at Virginia Tech
SPEAKER: John Tyson

ABSTRACT. In 2016 Virginia Tech opened a new undergraduate B.S. degree in Systems Biology. In this presentation, I will explain the basic goals, scope and curriculum of the degree, and some of our experiences so far in delivering content to the first two cohorts of the major. The core of the degree consists of the following courses: SYSB 2025-26: Introduction to Systems Biology (3C-3C) SYSB 3035-36: Systems Biology of Genes and Proteins (4C-4C) SYSB 3115-16: Network Dynamics & Cell Physiology (4C-4C) SYSB 4065-66: Research Experience in Systems Biology (4C-4C) SYSB 4135-36: Professional Development in Systems Biology (2C-2C) My intention is to stimulate discussion on the desirability and feasibility of undergraduate degree programs in systems biology at universities around the world.

10:40
Systems biology education – When to start?
SPEAKER: Thomas Höfer

ABSTRACT. Research in systems biology requires knowledge and skills that are traditionally taught in very different curricula (e.g., physics or mathematics versus biology). To address this problem, master's degree courses have been set up. By comparison, courses that teach the foundations of systems biology at the bachelor level have remained rare. Drawing on my experiences as a student of such a degree course and as a university teacher, I will argue that this direct path to systems biology has distinct advantages – for both students and professors.

10:50
Educational experiences in systems and synthetic biology: a mathematician’s perspective
SPEAKER: Brian Ingalls

ABSTRACT. In this talk I will relate my experiences teaching systems and synthetic biology to interdisciplinary audiences at the University of Waterloo. In 2004, I developed an undergraduate course in differential equation-based kinetic modelling for an audience from Science, Engineering, and Mathematics programs. The course presumes minimal mathematical prerequisites, focusing on development of models and interpretation of analysis, rather than on the details of analytic techniques. I worked my lecture notes into a textbook, which was published as “Mathematical Modelling in Systems Biology: an introduction”. Since 2005, I’ve had the opportunity to present these modelling techniques in the context of synthetic biology in workshops in Toronto and Alberta, to our to Waterloo iGEM teams, and in an undergrad Synthetic Biology Design course I co-teach. Mathematical modelling is often unfamiliar to life scientists working in systems and synthetic biology, and so communicating the potential — and the limitations — of these tools remains an important educational challenge.

11:00
TBA
SPEAKER: Edda Klipp
11:05
TBA
SPEAKER: Tomas Helikar
11:10
Discussion for Systems Biology Education
10:30-12:30 Session 21B: Parallel Session VII b: Regulatory Network II
Chair:
Location: Brush Mountain A & B
10:30
Analysis and classification of differential production within toxin-antitoxin systems using large datasets
SPEAKER: Heather Deter

ABSTRACT. Toxin-antitoxin (TA) systems are key regulators of bacterial persistence, a multidrug-tolerant state found in bacterial species that is a major contributing factor to the growing human health crisis of antibiotic resistance. Type II TA systems consist of two proteins, a toxin and an antitoxin; the toxin is neutralized when they form a complex. The ratio of antitoxin to toxin is significantly greater than 1.0 in the non-persister state, but this ratio is expected to become smaller during persistence. Analysis of multiple datasets (RNA-seq, ribosome profiling) and results from translation initiation rate calculators reveal multiple mechanisms that ensure a high antitoxin to toxin ratio in the non-persister state. The regulation mechanisms include both translational and transcriptional regulation. We classified E. coli type II TA systems into four distinct classes based on the mechanism of differential protein production between toxin and antitoxin. We find that the most common regulation mechanism is translational regulation. This classification scheme further refines our understanding of the fundamental mechanisms underlying bacterial persistence, especially regarding maintenance of the antitoxin to toxin ratio.

10:50
Exploring the dynamic regulation underlying synchronous chromosome splitting in anaphase
SPEAKER: Silke Hauf

ABSTRACT. When cells divide, they undergo an ordered series of events, called the cell cycle. The regulatory pathways that govern major transitions within the cell cycle typically contain positive feedback loops, which ensures that the transition occurs quickly and is irreversible. One important step within the cell cycle is anaphase: chromosomes split abruptly and synchronously to move into the emerging daughter cells, which is triggered by the protease separase. Any error in this step can lead to abnormal chromosome numbers in the daughter cells, a condition observed in cancer cells. Whether the abrupt and faithful splitting of chromosomes in anaphase necessitates positive feedback regulation is unclear. We systematically analyzed the dynamics of sister chromatid separation in fission yeast at the single-cell level. All chromosomes split during a narrow time window. Separase activity and the degradation kinetics of its inhibitor, securin, are the main determinants of this synchronicity. Mathematical modelling on the basis of our findings suggests that synchronicity is established in the absence of feedback regulation. Simple assumptions about securin-separase association and securin degradation are sufficient to explain rapid separase release and abrupt chromosome splitting. Hence, chromosome splitting, being already irreversible by nature, may be one of the few major cell cycle transitions that can operate without positive feedback.

11:10
Analysis of a dynamic model of guard cell signaling reveals the stability of signal propagation
SPEAKER: Xiao Gan

ABSTRACT. Analyzing the attractors of dynamic models of biological systems can provide valuable insight into biological phenotypes. To find out the best methods of attractor analysis of dynamical systems with a large state space, we identified the attractors of a multi-level, 70-node dynamic model of the stomatal opening process in plants, with state space of ~1031. We first reduced unregulated nodes and simple mediator nodes, and then simplified the regulatory functions of selected nodes. Next, we performed attractor analysis on the resulting 32-node reduced model by three methods. First, we converted the model into a Boolean model, then applied two attractor-finding algorithms, stable motif and GINsim. The stable motif based method computes all attractors of a Boolean model including all possible oscillations; GINsim computes all steady states and confirms oscillations by simulation. Second, we performed theoretical analysis of the regulatory functions, validating the attractor conclusions and extending the conclusions to perturbed cases. Third, we extended the stable motif method from Boolean to multi-level and apply it. The method is based on an expanded representation of the network that incorporates all the regulatory functions, by introducing virtual nodes for each state of an original node and composite nodes that express the ‘And’ Boolean operation. Stable motifs, subgraphs whose nodes’ states can stabilize on their own, can then be identified from the topology of the expanded network, leading to reduction of the network and identification of attractors (see Figure 1). Combining the results of all three methods, we concluded that all nodes except two in the reduced model have a single attractor; and only two nodes can admit oscillations. The multistability or oscillations of these four nodes do not affect the stomatal opening level in any situation. The high degree of attractor similarity shows stability of signal propagation, despite the complex structure of the network.

11:30
Identifying frequent patterns in biochemical reaction networks using graph-mining
SPEAKER: Ron Henkel

ABSTRACT. Background: It is common practice to represent molecular and cell biological processes as biochemical reaction networks using computational methods. Such models can be studied, analysed, and compared to improve our understanding of biology. One possible similarity measure for biological reaction networks is the occurrence of similar structures, including motifs, in the networks’ computational representation. The large number of models available from open model repositories, such as BioModels or the Physiome Model Repository, demands automated methods to support researchers in the identification of recurring patterns and in the recognition of biologically relevant motifs within a set of potentially interesting models. Specifically, for the problem of finding patterns in large, sparsely connected networks only partial solutions exist.

Result: We propose a workflow to identify frequent structural patterns in biochemical reaction networks encoded in the Systems Biology Markup Language (SBML, [1]). To detect such patterns, we implement a sub-graph mining algorithm. Once patterns are identified, the textual pattern description can automatically be converted into a graphical representation easily understandable for a user. Furthermore, information about the pattern distribution among the selected set of models can be retrieved. The workflow can be applied to a custom set of models or to models already existing in our graph database MaSyMoS [2]. The workflow was validated with 575 models from the curated branch of BioModels [3]. As a showcase, we highlight interesting and frequent structural patterns that appear within this model set. We further provide exemplary patterns that incorporate terms from the Systems Biology Ontology (SBO, [4]).

Conclusions: The occurrences of frequent patterns may give insight into central biological processes, evaluate biological motifs, or serve as a similarity measure for models that share common structures.

References [1] Hucka, M. et al. "The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models." Bioinformatics 19.4 (2003): 524-531. [2] Henkel, R. et al. "Combining computational models, semantic annotations and simulation experiments in a graph database." Database 2015 (2015): bau130. [3] Le Novère, N. et al. "BioModels Database: a free, centralized database of curated, published, quantitative kinetic models of biochemical and cellular systems." Nucleic acids research 34.suppl 1 (2006): D689-D691. [4] Courtot, M. et al. "Controlled vocabularies and semantics in systems biology." Molecular systems biology 7.1 (2011): 543.

11:50
Cell Cycle Control in Budding Yeast: Robustness of Model Predictions

ABSTRACT. We have recently crafted a new mathematical model of cell cycle progression in yeast that successfully explains the observed phenotypes of 257 budding yeast mutant strains. The model can be used to predict the phenotypes of novel combinations of mutant alleles and the number of such predictions is estimated to exceed the experimentally tested combinations by the order of 1.0e5. Thus, prioritizing predictions for experimental validation is an important step before starting expensive wet lab tests. We have developed an automated method to search for many alternative parameter sets that are consistent with experimental data and use those sets to explore the dependence of predictions on choice of parameters. We start with a “basal” parameter set that is consistent with 257 experimentally characterized mutants and search for alternative parameter sets by randomly perturbing all parameters simultaneously, retaining only those sets that are consistent with a selected set of 90 most informative mutant strains. We have found ~1000 alternative parameter sets for our model of cell cycle regulation, each parameter set can be thought as an alternative model that is also capable to explain a set of experimental data. Performing sensitivity analysis we have categorized selected predictions as robust if models with alternative parameter sets produce mostly the same results as the model with the “basal” set, or fragile otherwise. Furthermore, principal component analysis demonstrates tight fitting of the model to experimental data, contradicting the popular opinion that all systems biology models are universally “sloppy” in fitting data. Robust predictions depend on the regulatory network itself rather than specific parameter values and thus future experimental tests of robust predictions will either confirm the underlying molecular mechanism or provide new insights into how the cell division cycle is regulated. Fragile predictions are sensitive to values of adjustable parameters and thus their experimental characterization will be useful to constrain adjustable parameters of the model.

12:10
Evaluating the effects of measurement noise on the inference of biological regulatory networks using Modular Response Analysis

ABSTRACT. Modular Response Analysis (MRA) is a powerful mathematical framework, which allows to unravel the topology of intracellular regulatory networks from consecutive perturbation experiments. This method relies on steady state measurements of relevant species of the investigated network. MRA is widely used in systems biology, including topology estimation from Western blot data, which are known to be quite noisy. The calculation of connection coefficients describing the network’s interactions requires the solution of a reverse problem, which may be unstable with respect to noisy data.

In this study, we investigate how errors affect network inference using MRA and its robustness with respect to noise. Uncertainty in the input data propagates in a nonlinear manner to affect the variability of the inferred coefficients, producing possibly incorrectly inferred topologies. Through an extensive in silico study of models of two well-known biological systems, MAPK and p53, we compare different methodologies and computational approaches to handle noisy data in the inference process. We look at different perturbation strengths, different replicates’ numbers, and we solve the reverse problem using either ordinary or total least squares.

Our results demonstrate that it is not obvious which MRA formulation provides more accurate solutions, in terms of bias and variance of the inferred coefficients, and even small measurement noise can lead to erroneous results. Nonlinear dependencies of the equilibria of the system with respect to perturbed parameters can lead to biased estimators of the connection coefficients. Results differ for the two models, having different degrees of such nonlinearities. Nevertheless, the correct experimental design and MRA formulation maximise the probability of a correctly inferred topology.

The proposed analysis brings a deeper understanding of the effects of noisy data on regulatory networks inference via MRA, demonstrating that the choice of experimental design and MRA formulation is crucial for a reliable topology estimation.

10:30-12:30 Session 21C: Parallel Session VII c: Computational Methodology III
Location: Old Dominion Ballroom
10:30
E-cyanobacterium.org: A Web-based Platform for Systems Biology of Cyanobacteria
SPEAKER: Jan Červený

ABSTRACT. E-cyanobacterium.org is an online platform providing tools for public sharing, annotation, analysis, and visualization of dynamical models and wet-lab experiments related to cyanobacteria. The platform is unique in integrating abstract mathematical models with a precise consortium-agreed biochemical description provided in a rule-based formalism. The general aim is to stimulate collaboration between experimental and computational systems biologists to achieve better understanding cyanobacteria. We developed the tool that focuses on providing a general online platform for systems biology of cyanobacteria unifying the state-of-the-art knowledge base, related kinetic models and wet-lab experiments. In contrast to existing tools such as Biomodels.net or CellML, which provide general repositories for biological models, e-cyanobacteria.org is directly focused on cyanobacteria organisms. Our platform provides a unique solution based on integrating the well-acknowledged systems biology standards with advanced computer scientific techniques targeting the mentioned issues. Moreover, using in silico experiments for estimation of optimal culturing conditions and runtime multi-parameter optimization of variable environmental parameters can help to simplify optimization of real biotechnological applications, both on the side of cell biology and photobioreactors design. Current version is available at http://www.e-cyanobacterium.org/ and includes the following functions: – Biochemical Space – formal representation of elemental reactions facilitated by cyanobacteria biochemical entities, the representation is systematically organized by reflecting the hierarchy of biochemical processes ranging from the environment to the cell compartments and accompanied with Biochemical Space Language – rule-based formal language; – Computational Models – repository of stoichiometric and kinetic models providing simulation and static analysis; – SBML Compatibility – models projected onto the Biochemical Space can be exported into well-annotated SBML files; – Wet-lab Experiments – import and storage of time-series experiments, relation to models; – Annotation – detailed annotation of all system components reflecting annotation standards (OBO, OWL); – Content Visualization (Biochemical Space in means of process hierarchy schemas and modelling/experimental data). In the current version, the following processes of cyanobacteria are covered: environmental processes, respiration and photosynthesis, circadian clock and metabolism. Environmental processes focus on precise positioning of cyanobacteria into the context of its environment. Since the website primarily targets in vitro cultivation conditions in a bioreactor, we have compiled relevant elemental reactions. Processes of respiration and photosynthesis cover the energetic components of cyanobacteria. Circadian clock forms core of cyanobacteria and drives most important processes. Above these cellular processes, the metabolic part of the Biochemical Space forms a backbone that connects the bioenergetic components with metabolome and connects all key cellular processes with the general processes occurring in the environment. Currently available tools mostly do not provide sufficient means of supporting entire systems biology workflow. Especially, this applies to existing domain-specific tools devoted to cyanobacteria. Therefore, we believe that our web service makes a significant contribution. For future work, we plan to improve the mapping between mathematics and biology, to enhance the website with more analysis tools, and to automatize the comparison of models against experimental data. Moreover, biochemical space of cyanobacteria is continuously being extended and improved with interactive visualizations of reaction networks based on formal description provided in Biochemical Space Language.

10:50
Reverse-Engineering Gene regulatory Networks as Threshold Boolean Networks

ABSTRACT. Inferring the topology and dynamic of gene regulatory networks from data (time-course, input-output, or steady states) is one of the challenging problems in systems biology. Given time course experimental data, the objective is to identify the structure of the network as well as the rules of interaction among the genes of the network. However, even within the Boolean network framework, there usually are many Boolean networks that explain the available data. The so-called threshold Boolean networks(TBNs), initially developed to study neural networks, have been used to model a variety of gene regulatory networks. In a TBN, the future state of each node is determined based on a threshold and a linear combination of the current states of its neighbors. In this talk, we present an algorithm for identifying all threshold network models that reproduce a given Boolean dataset. Our method is rooted in algebraic combinatorics.

11:10
GEF-H1 regulation by microtubule dynamics unveiled by fluctuation analysis of biosensor images
SPEAKER: Jungsik Noh

ABSTRACT. RhoA GTPase is a signaling molecule implicated in orchestrating cytoskeleton dynamic processes such as actin assembly and actomyosin contractility. Its activity is regulated by several guanine exchange factors, of which one, GEF-H1, is thought to localize in a non-activating state to microtubules (MTs). By acute pharmacological induction of MT disassembly across the entire cell it was shown that RhoA is activated in a GEF-H1 dependent fashion. This suggests that GEF-H1 mediates an important link between MT and actin network dynamics. However, these previous whole-cell perturbation studies have left unanswered whether local MT dynamics drives changes in GEF-H1 activation. This would establish a pathway for MT-actin regulatory cross-talk at the micron and second length and time scales, respectively. To test this possibility we analyzed the subcellular dynamics of GEF-H1 activity in response to MT disassembly events using the spontaneous fluctuation in GEF-H1 activity imaged by a novel biosensor design in conjunction with MT +TIP particle tracking of MT growth. We developed a novel statistical framework to test the hypothesis that the discrete and sparse MT disassembly events modulate the continuous GEF-H1 signal. Overall, GEF-H1 biosensor images exhibited higher activity near the cell edge and shadows along the shafts of MTs confirming that GEF-H1 sequestration by MTs indeed causes deactivation, whereas release from disassembling MTs at the cell periphery may trigger GEF-H1 activation. Indeed, we found that subcellular signals of GEF-H1 activity tend to increase after MT disassembly events with a delay of about 10 sec in breast cancer cells. Our analysis also suggested that MT assembly deactivated GEF-H1 along growth trajectory. On analytical side, to our knowledge this work is the first to statistically couple discrete molecular events to the continuous output of a molecular population.

11:30
Segmenting four-dimensional fluorescence microscopic image using Convolutional Neural Network
SPEAKER: Yuta Tokuoka

ABSTRACT. To segment microscopic images is one of the important tasks for quantitative analysis of biological phenomena. The problem of segmentation in biological imaging is that the parameter value of each image processing should be determined beforehand, and the parameter value depends on the feature of each image and imaging conditions. Recently, many studies have been reported which applied Convolutional Neural Network (CNN) for segmenting two-dimensional microscopic image to solve this problem. However, because it is common to acquire multi dimensional microscopic images along with the development of imaging technology, proposals of segmenting multi dimensional microscopic images are desired. In this study, we propose the segmentation algorithm for four-dimensional fluorescence microscopic images using CNN. We implemented an algorithm to segment nuclei from four-dimensional fluorescence microscopic images of mouse embryos using CNN. In dataset creation, we cropped one time point in the three-dimensional fluorescence microscopic image of 4 cell stage of the mouse embryo, and the Ground Truth was created manually. We also interpolated the microscopic image and the Ground Truth in order to match the actual scale. In the implementation of the learning model, segmentation algorithm was implemented based on CNN. Four-dimensional fluorescence microscopic images of 2 to 14 cell stages were used for the evaluation of segmentation. With the current implementation, the voxel accuracy of segmentation achieved 98.74% in average. As a future prospect, we will acquire quantitative indexes from four-dimensional fluorescence microscopic images of early-stage embryos by a further improved segmentation algorithm, and select the most effective indexes to evaluate the quality of embryos.

11:50
CrossPlan: Systematic Planning of Genetic Crosses to Validate Mathematical Models

ABSTRACT. Mathematical models of cellular processes can systematically predict the phenotypes of novel combinations of multi-gene mutations. Searching for informative mutants and prioritizing them for experimental validation is challenging since the number of possible combinations grows explosively. Moreover, keeping track of the genetic crosses needed to make new mutants and planning sequences of experiments is unmanageable when the experimenter is deluged by hundreds of potentially informative predictions to test.

We present CrossPlan, an algorithm for systematically planning genetic crosses to make a set of target mutants from a set of source mutants. We base our approach on a generic experimental workflow used in performing genetic crosses in budding yeast. CrossPlan uses an integer-linear-program (ILP) to maximize the number of target mutants that we can make under certain experimental constraints.

We apply our method to a comprehensive mathematical model of the protein regulatory network controlling cell division in budding yeast. The number of target mutants we can plan increases linearly with the number of batches planned. We also extend our solution to incorporate other experimental conditions such as a delay factor that decides the availability of a mutant and genetic markers to confirm gene deletions. Our analyses reveal that incorporating the requirement that each gene deletion be associated with a unique marker has only a marginal effect on the number of planned mutants. Our method outperforms a greedy method that plans batches one at a time. Interestingly, planning two or three batches at a time is nearly as optimal as planning all batches simultaneously. The experimental flow that underlies our work is quite generic and our ILP-based algorithm is easy to modify. Hence our framework should be relevant in mammalian systems as well.

12:30-14:00 Session : Friday Lunch
Location: Commonwealth Ballroom