in room "Poster 1" on gather.town

2 - Predicted ‘wiring landscape’ of Ras network in 29 human tissues
Simona Catozzi and Christina Kiel

4 - Optogenetic cell signaling in plants and using phase-separation – modelling two frontiers in synthetic biology
Franz-Georg Wieland, Matias D. Zurbriggen, Wilfried Weber and Jens Timmer

6 - Periodic propagating activation waves coordinate Rho GTPase network dynamics at the leading and trailing edges during cell migration
Oleksii Rukhlenko, Alfonso Bolado-Carrancio, Elena Nikonova, Mikhail Tsyganov, Anne Wheeler, Amaya Garcia Munoz, Walter Kolch, Alex von Kriegsheim and Boris Kholodenko

8 - Molecular Noise-Filtering by Phospholamban Pentamers
Daniel Koch

10 - Reversion of Cisplatin Chemoresistance in a Triple Negative Breast Cancer Subtype
Lauren Marazzi, Pooja Kumar, Edison T Liu and Paola Vera-Licona

12 - Towards integrative mechanistic models of mammalian cell responses to extracellular perturbations: growth factors, hormones, and cytokines
Cemal Erdem, Sean M. Gross, Laura M. Heiser and Marc R. Birtwistle

14 - Integration of the cardiopulmonary receptors in a mathematical model of the response to the Valsalva Maneuver
Martin Miranda and María Rodríguez

16 - Modelling Proliferation on Cancer Pathways
Simon Merkt, Julio R. Banga, Leonard Schmiester, Paul Stapor, Daniel Weindl and Jan Hasenauer

18 - A Novel Mathematical Model of Cholesterol Metabolism and its Intersection with Atherosclerosis
Mark Tomás Mc Auley and Callum Davies

20 - Specification of Treatment and Intervention Timecourses alongside SBML and PEtab
Dilan Pathirana, Elba Raimundez, Leonard Schmiester, Paul Stapor, Daniel Weindl and Jan Hasenauer

22 - CobraMod: A pathway‑centric curation tool for constraint‑based metabolic models
Stefano Camborda La Cruz and Nadine Töpfer

24 - A Wall-time Minimizing Parallelization Strategy for Sequential Approximate Bayesian Computation
Felipe Reck, Yannik Schälte, Emad Alamoudi and Jan Hasenauer

26 - Benchmarking optimization algorithms for solving ODE-constrained parameter estimation problems in systems biology
Leonard Schmiester, Fabian Fröhlich, Daniel Weindl and Jan Hasenauer

28 - pyPESTO - a python Parameter EStimation TOolbox
Paul Jost, Jakob Vanhoefer, Paul Stapor, Yannik Schälte, Jan Hasenauer

30 - A systematic workflow to assess the useability of data in model development
Tara Hameed and Reiko Tanaka

---

Simona Catozzi and Christina Kiel

Ras is an important hub protein at the head of numerous signaling pathways and plays a role in various types of cancers, notably pancreas, colon and lung. The usual suspects are three oncogenic isoforms - i.e. HRAS, KRAS and NRAS - that are highly mutated and drive tumorigenesis [1]. Our study is based on the paradigm of network medicine that sees disease as a perturbation of a network of interconnected proteins orchestrating cell's physiology and phenotype through signal transduction. As such, we built a mechanistic model of the interactions of the three Ras oncoproteins with their direct interactors (effectors), with protein abundances and binding affinities being the system's parameters, in order to study elementary patho-/physiological conditions of Ras network [2].

Using high-quality proteomic data from 29 (healthy) human tissues [3], we quantified the amount of individual Ras-effector complexes, and characterized the (stationary, reference) Ras “wiring landscape” specific to each tissue. We simulated mutant- and stimulus-induced network re-configurations, miming respectively cancerous and physiological state, and compared them to the reference network. Moreover, we investigated the contribution of the input parameters (binding affinities and effector concentrations) in determining the complex formations underlying the specific wiring landscape, by 3D data interpolation onto (tissue-specific) surfaces. This revealed that high affinity - more than high concentration, - is critical for complex formation. As a consequence, we analyzed local and global binding affinity fluctuations and assessed their impact on the system's robustness. Further research will aim at the calibration of the binding affinities, based on the Ras-effector complexes and the activation of the associated downstream pathway.

[1] Hobbs G.A. et al., J. Cell Sci 129, 1287–1292 (2016).

[2] Ibáňez Gaspar V. et al., Small GTPases (2020).

[3] Wang D. et al., Mol Syst Biol 15(2), e8503 (2019).

Franz-Georg Wieland, Matias D. Zurbriggen, Wilfried Weber and Jens Timmer

Controlling cellular processes with light by introducing gene switches sensitive to different light conditions allows spatio-temporally precise manipulation of these processes. We have introduced these optogenetic switches in two recent applications. PULSE, the plant-usable light-switch elements allows for the use of optogenetic switches in plants under ambient white light, which has so far not been possible. In a further study, we showed how phase separation in the cell through intrinsically disordered regions can increase gene expression many-fold. For both applications, we created an ordinary differential equation model to accurately describe the system, assess its uncertainties, make accurate predictions, and to validate the results with independent measurements. The mathematical modelling allowed us to quantitatively characterize PULSE in protoplasts and to predict the systems behaviour for a wide range of stimulations. It furthermore allowed us to assess the effect of phase-separation on gene expression and the resulting fold-change of the transcription activation for two distinctly different systems with a joint approach.

1. Ochoa-Fernandez R, Abel NB, Wieland F-G, Schlegel J, Koch L-A, Miller JB, Engesser R, Giuriani G, Brandl SM, Timmer J, et al.: Optogenetic control of gene expression in plants in the presence of ambient white light. Nat Methods 2020, 17:717–725.

2. Schneider N, Wieland F-G, Kong D, Fischer AAM, Hörner M, Timmer J, Ye H, Weber W: Liquid-liquid phase separation of light-inducible transcription factors increases transcription activation in mammalian cells and mice. Sci Adv 2021, 7:eabd3568.

Oleksii Rukhlenko, Alfonso Bolado-Carrancio, Elena Nikonova, Mikhail Tsyganov, Anne Wheeler, Amaya Garcia Munoz, Walter Kolch, Alex von Kriegsheim and Boris Kholodenko

Cell migration relies on the coordination of actin dynamics at the leading and the trailing edges. During the mesenchymal type of migration, protrusive filamentous actin (F-actin) is cyclically polymerised/depolymerised at the cell’s leading edge, whereas the contractile, actomyosin-enriched trailing edge forms the rear. How these different dynamics coexist and get coordinated within the cell is unclear.

We have elucidated the spatial profiles of RhoA-Rac1 interactions in motile MDA-MB-231 breast cancer cells. Using proximity ligation assays, we showed that the concentration of complexes formed by RhoA and its downstream effectors DIA and ROCK depends on the spatial location along the longitudinal axis of polarized cells. RhoA primarily interacts with DIA at the cell leading edge, whereas RhoA - ROCK interactions are the strongest at the cell rear. Based on these findings, we have built a mathematical model to analyse RhoA-Rac1 signalling in space and time. The model predicted and the experiments confirmed that at the cell front the GTPase network exhibits oscillatory behavior with high average Rac1-GTP, whereas at the cell rear there is a (quasi)steady state with high RhoA-GTP and low Rac. The front and rear are connected by periodic, propagating GTPase waves. When the wave reaches the rear, RhoA-GTP transiently oscillates and then, following the rear retraction, the GTPase network dynamic pattern returns to the original state. Our modelling and experimental results showed how different GTPase dynamics at the leading edge and the trailing edge can govern distinct cytoskeleton processes and how moving cells reconcile these different dynamics. In addition, the model correctly described network responses to different inhibitor perturbations. The RhoA-Rac1 interaction network model defines minimal, autonomous biochemical machinery that is necessary and sufficient for biologically observed modes of cell movement.

Daniel Koch

Cardiac relaxation depends on the reuptake of calcium into the sarcoplasmic reticulum by the calcium pump SERCA. Under resting conditions, SERCA is inhibited by monomers of the micropeptide phospholamban (PLN). During the “fight-or-flight” response, β-adrenergic stimulation leads to phosphorylation of PLN by protein kinase A (PKA), thereby releasing SERCA inhibition and promoting fast calcium reuptake. One of the open questions in the field is the physiological role of PLN homo-pentamers which are mainly regarded as an inactive storage or buffering form for monomeric PLN. Based on mathematical modeling and biochemical experiments, we present evidence that PLN pentamers can delay phosphorylation of PLN monomers via substrate competition. Simulations and model analyses further show that phosphorylation of PLN may be ultrasensitive and bistable due to cooperative dephosphorylation of pentamers. Both effects enhance the noise-filtering behavior of the system in numerical simulations. This indicates that PLN pentamers may act as molecular noise-filters which could ensure consistent monomer phosphorylation and SERCA activity despite fluctuating PKA-activity in the upstream signaling network. Importantly, these results offer novel perspectives on the role of PLN in cardiac arrhythmias resulting e.g. from pathogenic PLN mutations such as R14del.

Lauren Marazzi, Pooja Kumar, Edison T Liu and Paola Vera-Licona

Triple Negative Breast Cancer (TNBC) represents a diverse group of cancers with a high prevalence among African American, Hispanic ethnic groups, and a younger age of onset compared to other breast cancer subtypes. The Tandem Duplicator Phenotype (TDP) is a genomic configuration characterized by numerous distributed tandem duplications found in 40% of TNBCs. A subset of TDP tumors are initially highly sensitive to the chemotherapy drug cisplatin, but they eventually develop resistance. Development of acquired drug resistance can be viewed as a systems-level process driven by a core resistance intracellular signaling network, where corresponding long-term dynamic behaviors (attractors) can be associated to the resistant and sensitive phenotypes of the tumor. The reprogramming of a cancer cell’s long-term behavior away from the resistance phenotype may enhance a tumor’s sensitivity to treatment. This reprogramming can be achieved through interventions on the control targets of resistance-associated phenotypes. This project aims to take a dynamical systems approach to identifying combinations of therapeutic targets for cisplatin resistance reversion in a TDP tumor in silico.

We have developed a pipeline for constructing a static intracellular signaling network with multi-omics data from a TDP TNBC Patient Derived Xenograft (PDX) model treated with cisplatin. The resultant network captures known TNBC dysregulated genes and cisplatin resistance-associated genes. Control targets of the network were identified using structure-based control theory. Combinations of resistance-reversion control target perturbations were estimated using signal flow analysis. These intervention target predictions will be validated experimentally to identify chemotherapy adjuvants for the TNBC TDP subtype.

Cemal Erdem, Sean M. Gross, Laura M. Heiser and Marc R. Birtwistle

A key missing capability in current cancer research is the ability to predict how a particular single cancer cell will respond to microenvironmental cues or a drug cocktail. Yet, it is not even possible to perform this task well for normal healthy cells. This work builds on the hypothesis that first principles, mechanistic models of how cells respond to different perturbagens will ultimately improve drug combination response predictions. However, building such single-cell models of complex, large-scale, and incompletely understood systems remains an extremely challenging task. Here, we defined an open-source pipeline for scalable, single-cell mechanistic modeling that converts simple, annotated input files (structured lists of species, parameters, reaction types) into an SBML model file. Using this pipeline, (i) we re-created one of the largest pan-cancer signaling models in the literature (774 species, 141 genes, 2400 reactions), (ii) enlarged the model to include Interferon-γ (IFNγ) signaling pathway (950 species, 150 genes, 2500 reactions), and (iii) re-parametrized the model to test and prioritize candidate mechanisms. Specifically, we used the enlarged model to test alternative mechanistic hypotheses for the experimental observations that IFNγ inhibits epidermal growth factor (EGF)-induced cell proliferation. Our single-cell-simulation-based analysis suggested, and experiments support that these observations are better explained by IFNγ-induced SOCS1 expression sequestering activated EGF receptors, thereby downregulating AKT activity, as opposed to direct IFNγ-induced upregulation of p21 expression. Finally, our new modeling format is available online and compatible with high performance (Kubernetes) computing platforms, enabling us to study virtual cell population responses. Overall, our new model enables easy modification of large mechanistic models and simulation of thousands of single-cell responses to multiple ligands and drug combinations.

Martin Miranda and María Rodríguez

The control of the autonomic nervous system over the cardiovascular system provides the organism with the ability to rapidly adapt to external stimuli, such as postural changes. The main short-term blood pressure control mechanism is exerted by high-pressure baroreceptors located in the aorta and carotid arteries. Recently, great interest has been put into the modulation of cardiopulmonary receptors (or low-pressure baroreceptors) on the muscle sympathetic neural activity to understand its effects on the cardiovascular response. However, direct measurements require invasive methods that are impractical in humans, whereby indirect evaluations such as the Valsalva Maneuver are currently used. Mathematical models have been used in many fields to understand complex systems. In this work, the delay differential equation (DDE) model presented in Randall et al. (2019) was extended to include the effect of modulation of cardiopulmonary baroreceptors.

The cardiovascular response (ECG and blood pressure) to the Valsalva Maneuver of six healthy participants was recorded using a Finapres device in the laboratory of the Institute of Biological and Medical Engineering of the Pontificia Universidad Catolica de Chile. Patient-specific parameters were obtained by fitting the output of the mathematical model to the signals of each participant. Finally, a local sensibility analysis was performed to identify the most relevant parameters. The extended model allowed a better understanding of how cardiopulmonary baroreceptors modulate the sympathetic output by studying the heart rate response. The inclusion of microneurography recordings will allow modeling the system taking as main output a direct measure of the sympathetic nerve activity, which will permit drawing better conclusions.

Simon Merkt, Julio R. Banga, Leonard Schmiester, Paul Stapor, Daniel Weindl and Jan Hasenauer

Testing potential cancer inhibitors on mechanistic models of signalling pathways can substantially speed up the development of new tumor suppressors. However, if one considers models describing multiple pathways, mapping proteins which promote or restrain cancer growth to a specific cell count number can be a challenge. An intuitive approach would be modelling cell division exponentially with a rate described by the fraction of weighted sums of the inhibitors and activators of growth. But during fitting of these weights to experimentally obtained measurements we encountered serious numerical instability.

To tackle this we investigated various combinations of proliferation models and state transformations. We tested those on a toy model with 17 states and 20 reactions as well as on a large scale model with over 3000 states and 5000 reactions using artificial and experimental data, respectively. Our study suggests that simulating the logarithm of a cell count as state variable can substantially improve the numerical properties of the proliferation model. However, which of the proliferation models is suited best for describing cancer growth and its inhibition is still subject to to further research.

Mark Tomás Mc Auley and Callum Davies

Atherosclerotic cardiovascular disease (ACVD) is the leading cause of morbidity and mortality amongst Western populations. Dysregulated cholesterol metabolism is the principal risk factor for ACVD. Cholesterol metabolism is maintained by an array of complex interacting biological mechanisms. These regulatory processes are multicomponent in nature and display emergent properties. Likewise, the mechanisms which underpin atherogenesis are nontrivial and interact across time and space. Mathematical modelling, which is a core constituent of the systems biology paradigm has played a pivotal role in deciphering the dynamics of both cholesterol metabolism and atherosclerosis. However, to date no model has successfully integrated cholesterol biosynthesis, whole body metabolism and atherosclerosis. In this work we describe how we merged a model of cholesterol metabolism with a model of atherosclerosis. The SBML encased models were merged in COPASI, which was then utilised to reparametrize and simulate the combined model. Using our novel in silico system we were able to reproduce output from the parent models. As a result, we have created a more complete theoretical representation of cholesterol metabolism and its intersection with atherosclerosis. It is hoped this new system will help us to develop a deeper understanding of the relationship between cholesterol metabolism and ACVD.

Dilan Pathirana, Elba Raimundez, Leonard Schmiester, Paul Stapor, Daniel Weindl and Jan Hasenauer

When modeling patient-specific treatments or epidemic interventions, simulations can involve parameters that are fixed to different values as simulated time progresses. Examples include the adjustment of the rate of intravenous infusion of glucose into a patient's bloodstream over time, or the severity of a government-mandated lockdown during an epidemic, with assumed effects on model parameters, for example disease transmission rates, as the severity changes over time.

Models can involve hundreds of such parameter value changes over the course of a single simulation. In SBML, current options for the specification of such a parameter timecourse include events and piecewise functions. However, these options directly encode the timecourse information into the model. This can yield an unwieldy model with long import times prior to simulation, and reduces the portability of the model to different scenarios, such as patient-specific treatment timecourses or adjusted epidemic intervention strategies.

The alternative presented here includes a simple file format, such that these timecourses can be specified independently of the model, and a Python package, which can be used to simulate and optimize models with parameter timecourses using pre-existing tools. The format and package is designed for use with pre-existing SBML models and PEtab problems. Example use cases are also presented.

Stefano Camborda La Cruz and Nadine Töpfer

Genome-scale metabolic models (GEMs) and their analysis by constraint‑based metabolic modeling techniques are a popular tool to study metabolic systems at a large‑scale. Several software tools for Constraint‑Based Reconstruction and Analysis (COBRA) are available such as COBRApy(1) which is based on the popular programming language Python. These tools offer a wide range of functionalities for model modification and metabolic flux predictions. However, they all require manual addition of biochemical information, such as metabolites, reactions, and gene IDs which is time-consuming and error‑prone. Here we present CobraMod, a COBRApy extension for pathway‑centric modification and curation of GEMs. The open‑source package enables extending GEMs with biochemical data from various databases such as BiGG, BioCyc, and KEGG. Our tool automatically identifies and transforms biochemical information into the corresponding sets of metabolites and reactions. These sets can either be separately added or included as groups, denominated pathways. CobraMod curates each new pathway taking into consideration chemical formulas, duplicate elements, reversibility and mass balance of reactions, the capability to carry non‑zero fluxes for newly added reactions, track changes, and can add extra databases for data retrieval. Our package integrates the software package Escher(2) for visualizing pathways and their corresponding flux distributions and thus enables a pathway‑centric and user‑friendly analysis of the model. CobraMod uses Python standards and aims for stability, uniformity and speed. We exemplify these functionalities in a case study where we calculate the additional metabolic cost for synthesizing new metabolites in Arabidopsis thaliana.

1. Ebrahim, et al. (2013) BMC Syst Biol 7, 74

2. King ZA, et al. (2015) PLOS Comp Bio 11(8)

Felipe Reck, Yannik Schälte, Emad Alamoudi and Jan Hasenauer

We present an unbiased wall-time minimizing parallelization strategy for Sequential Monte Carlo samplers (ABC-SMC), by proactively starting sampling for the next iteration already, as soon as cores become idle.

The proactively sampled particles are based on a possibly biased preliminary population but are afterwards reweighted to account for said bias.

We evaluate this look-ahead sampling algorithm on toy models of various types and properties and on different recently published application models.

Further, we demonstrate that the algorithm can be combined with self-tuning (e.g., the adaptive selection of thresholds, population sizes and distance functions), which is essential for applications. Depending on the scenario, we observe an overall speed-up of up to 40\% compared to established approaches.

Hence, the proposed algorithm can improve the cost and runtime efficiency of ABC-SMC methods on high-performance infrastructure considerably.

Leonard Schmiester, Fabian Fröhlich, Daniel Weindl and Jan Hasenauer

Mechanistic models usually comprise unknown parameters which have to be inferred from experimental data. This is commonly done by optimizing an objective function, which can be a challenging task due to large computational requirements and ill-conditioned optimization problems. A multitude of optimization methods have been developed that implement different schemes to iteratively update the parameters. These optimization algorithms can differ substantially in performance and the identification of well-working methods is crucial for successful parameter estimation.

Here, we benchmark a large range of different optimization algorithms interfaced in the pyPESTO toolbox on more than 20 parameter estimation problems in systems biology. These problems are based on published models using real data to guarantee a realistic setting. The comprehensive collection of test models allows to develop guidelines for good optimizer choices. We highlight optimizers that perform particularly well and that could be used as default algorithms. The parameter estimation problems are implemented in the PEtab format which allows for easy extension of this study to additional optimizers implemented in other toolboxes.

Paul Jost, Jakob Vanhoefer, Paul Stapor, Yannik Schälte, Jan Hasenauer

Solving inverse problems and estimating parameters is a common task in systems and computational biology. Especially optimization and uncertainty analysis techniques play an important role when it comes to developing predictive computational models. Here, we present pyPESTO, a highly modular and efficient Parameter EStimation Toolbox, written in python, for systems biology problems. It is particularly suited for ordinary differential equations models specified in SBML and Petab and also usable for large-scale models by employing parallelization techniques and being tightly integrated with AMICI. In terms of uncertainty analysis, it can perform profile likelihoods computation, MCMC sampling, and includes recently developed ensemble techniques. PyPESTO comes with many convenience routines for visualization as well as for storing and reading of parameter estimation results.

Tara Hameed and Reiko Tanaka

Data sparsity is one of the bottlenecks we often encounter in model development, especially for disease modelling or in fields where interdisciplinary cross-collaboration is still being developed [1]. When a model is fit to sparse data, it is hard to discern whether potential model misfit is caused by inherent model misspecification, which requires reformulation of the model, or by data sparsity, which requires further data collection. We proposed a systematic workflow to assess the degree to which the available data can inform mathematical models theoretically, by upcycling a known statistical workflow that uses simulation studies [2]. The proposed workflow quantifies the useability of the experimental data in terms of expected parameter identifiability and model prediction. Application of the workflow to our mathematical model of early-stage invasive aspergillosis (pulmonary fungal infection), adapted from a previous model [3], allowed us to suggest future experiments that could provide more “useable” data to infer the model’s nonlinear interaction parameters and to make better predictions. The presented workflow could be useful when models are developed with data sparsity as a limiting factor for model-based inference.

[1] Cvijovic, M. et al. Bridging the gaps in systems biology. Molecular genetics and genomics : MGG vol. 289 727–734 (2014).

[2] Morris, T. P., White, I. R. & Crowther, M. J. Using simulation studies to evaluate statistical methods. Statistics in Medicine vol. 38 (2019).

[3] Tanaka, R. J. et al. In silico modeling of spore inhalation reveals fungal persistence following low dose exposure. Sci. Rep. 5, 13958 (2015).