Effective immune-cell responses depend on collective decision-making mediated by diffusible intercellular signaling proteins called cytokines. Here, we designed a three-dimensional spatio-temporal modeling framework and a precise finite-element simulation setup, to systematically investigate the origin and consequences of spatially inhomogeneous cytokine distributions in lymph nodes. We found that such inhomogeneities are critical for effective paracrine signaling, and they do not arise by diffusion and uptake alone, but rather depend on properties of the cell population such as an all-or-none behavior of cytokine secreting cells. Furthermore, we assessed the regulatory properties of negative and positive feedback in combination with diffusion-limited signaling dynamics, and we derived statistical quantities to characterize the spatio-temporal signaling landscape in the context of specific tissue architectures. Overall, our simulations highlight the complex spatiotemporal dynamics imposed by cell-cell signaling with diffusible ligands, which entails a large potential for fine-tuned biological control especially if combined with feedback mechanisms.

In this lecture I will show my lab's ongoing computational work in the parameterization and application of coarse-grain models to simulate peptidoglycan. While diving into details of structural biology of glycans and advanced techniques in coarse-grain parameterization, I will describe how this journey started several years ago, the unexpected findings we have observed -- with important consequence to our understanding of peptidoglycan's biological function -- and how we have tackled the challenges of representing this macromolecule within the Martini coarse-grain framework, so that we can eventually simulate bacterium-sized cell walls.

Evolutionary history, captured in the form of gene and species trees, informs every area of modern biology. However, distentangling gene and species evolution requires complex evolutionary models able to encompass various processes such as, gene duplications, gene losses, horizontal gene transfers, incomplete lineage sorting, and hybridization. Furthermore, analyzing the large amount of available genomic data under those models poses formidable computational challenges. In this talk, I will describe cutting-edge phylogenetic methods employed to unravel the intricate history of gene and species evolution. I will also outline their limitations and present promising avenues to surmount those challenges.

The cell is the simplest independent functional unit of life. However, even unicellular model organisms that are touted for their tractability are complex, possessing thousands of genes and proteins, many of
which remain uncharacterized even after decades of in-depth investigation. The quest for the simplest organism has been aided by advances in synthetic biology, which involve the redesign or novel construction of biological parts and modules. Using such tools, a minimal cell was constructed with a genome derived from the bacterium Mycoplasma mycoides, which contains only the smallest set of genes required for autonomous cellular life. I will discuss experiments used to test whether genome minimization constrained cellular evolution. Mutation rates were the highest among all reported bacteria, but were not affected by genome minimization. And while synthetic streamlining of the genome was initially costly, the minimal cell rapidly regained fitness in just 2000 generations of evolution. The only apparent constraint associated with genome minimization involved the evolution of cell size. Our findings demonstrate that natural selection can rapidly increase the fitness of one of the simplest autonomously growing organisms. Understanding how species with small genomes overcome evolutionary challenges provides critical insights into the persistence of host-associated endosymbionts, the stability of streamlined chassis for biotechnology, and the targeted refinement of synthetically engineered cells.

Throughout the tree of life populations have evolved the capacity to
contend with sub-optimal conditions by engaging in dormancy, whereby
individuals enter a reversible state of vanishing metabolic activity. The
resulting "seed banks" serve as long-lived reservoirs of genotypic and
phenotypic diversity. Of particular relevance is the case of microbial dormancy,
which has a fundamental impact on the evolutionary, ecological and also 
pathogenic character of biological communities.

However, despite its ubiquity in nature, dormancy is a rather 
new paradigm in stochastic interacting particle systems.
Here, it leads to novel effects, in particular based on the
introduction of memory, resilience and diversity into the underlying
systems. The resulting probabilistic structures, arising on different scales, 
are surprisingly rich, already when considering simple `toy models’, 
and lead to new universal behaviour.

In this talk, after providing some background on dormancy, 
I will present and discuss recent models for dormancy and seed banks 
in population genetics, evolution and ecology. Along the line, I will indicate topics for future mathematical and interdisciplinary research.

Many fields of science make extensive use of mechanistic forward models which are implemented through numerical simulators. Simulation-based inference aims to make it possible to perform Bayesian inference on such models by only using model-simulations, but not requiring access to likelihood evaluations. I will talk about our recent work on developing simulation based inference methods using flexible density estimators parameterised with neural networks, and applications to modelling problems in astrophysics, neuroscience and computational imaging.

Data-driven modeling and deep learning present powerful tools that are opening up new paradigms and opportunities in the understanding, discovery, and design of soft and biological materials. I will describe our recent applications of deep representational learning to expose the sequence-function relationship within homologous protein families and to use these principles for the data-driven design and experimental testing of synthetic proteins with elevated function. I will then describe an approach based on denoising diffusion models to restore all-atom resolution to coarse-grained molecular dynamics simulations of proteins to help break the time scale barrier in biomolecular simulations.

Mathematical modelling has long been used to understand embryological phenomena, and perhaps nowhere more successfully than in the study of pattern formation. In this talk I will briefly review some of my favourite models on pattern formation focusing on the insight that they’ve brought about, and paying special attention to the assumptions and abstractions that were made constructing them. One prevailing assumption in our models of pattern formation to date is that the role of cell movements is not generative and can be safely ignored, even though molecular patterns are often established in tissues undergoing morphogenesis where extensive cell rearrangements are taking place. I will argue that in order to understand the formation of molecular patterns and importantly, their evolvability, moving forward we should take the role of cell movements into account explicitly. I will finish outlining a new methodology for reverse-engineering GRNs driving pattern formation in developing tissues with extensive cell movements.

Advanced statistical methods are rapidly impregnating many scientific fields, offering new perspectives on long-standing problems. In materials science, data-driven methods are already bearing fruit in various disciplines, such as hard condensed matter or inorganic chemistry, while comparatively little has happened in soft matter. I will describe how we use multiscale simulations to leverage data-driven methods in soft matter. We aim at establishing structure-property relationships for complex thermodynamic processes across the

chemical space of small molecules. Akin to screening experiments, we devise a high-throughput coarse-grained simulation framework. Coarse-graining is an appealing screening strategy for two main reasons: it significantly reduces the size of chemical space and it can suggest a low-dimensional representation of the structure-property relationship. I will briefly mention a biological application of our methodology that led to the discovery of in vivo active compounds. Finally, I will mention a number of ways machine learning can help fulfill the promise of connecting models at different scales.

The spread of a pathogen within an infected host, as well as the development of counteracting immune responses represent complex dynamical systems that involve various cell types and factors. Identifying the key processes that determine the outcome of an infection or the generation of successful immune responses requires to decipher the interaction of individual cells and molecular factors across multiple scales in space and time. While advances within experimental techniques and high-throughput data acquisition have increased our ability to study immune processes in an unprecedented level of detail, analysing these data and getting to a mechanistic understanding of the various biological processes involved requires an interdisciplinary approach that combines image analysis, mathematical modelling and improved methods for parameter inference. In my talk, I will show how combining experimental data with data-driven modelling can be used to reveal the interactive dynamics of infection and immunity within tissues, and to infer the development of immunity in response to vaccination.

The Martini model is among the most popular coarse-grained (CG) models in the field of biomolecular simulation, due to its easy-to-use building block principle. Martini relies on amapping of two to four non-hydrogen atoms and has been parameterized using a top-down approach with thermodynamic data as the main target. With its growing use, however, several shortcomings have become apparent. These limitations pushed us to fully reparametrize the CG beads - the fundamental building blocks of Martini - largely untouched since 2007. In this talk, I will revisit the new Martini 3 model, showing its key features and some of the future developments which are coming. I will also show and discuss applications of the Martini 3 model in the design of drug and delivery systems, including the current challenges to routinely employe the model in academic and industrial discovery pipelines.

Spontaneous generation of patterns and structures occurs in many living systems and is linked to biological form and function. Such processes often take place on domains which themselves evolve in time, and they can be guided by or coupled to geometrical features. Using different biophysical examples, I will discuss how geometry affects spatial organization in cellular and multicellular systems. In particular, I will show recent work on the mechanics of embryo-uterus interactions during implantation [1], the spontaneous rotation of cell-cell interfaces in developing sensory cells [2], and the patterning of feather follicles [3].

The ability to understand and control biological systems at the molecular level is a critical goal in biology. However, predicting the behavior of biomolecules in complex environments such as living cells is a significant challenge. Moreover, biomolecular systems undergo rare yet essential conformational changes, such as protein folding and unfolding or DNA hybridization. While molecular dynamics simulation combined with statistical mechanics can offer insight into these processes, the large size of the systems and the long times involved make it challenging to fulfil this promise. Multiscale modeling is a potential solution, allowing for the development of hierarchical levels of description for different length and time scales in a biophysical system. I present two approaches that enable to bridge these length and time scales. The first approach, Transition Path Sampling (TPS), focuses on rare events such as protein unfolding and utilizes importance sampling to harvest an ensemble of trajectories between stable states. The second approach, MD-GFRD, simulates proteins using Molecular Dynamics at the microscopic level and Green’s Function Reaction Dynamics at the mesoscopic level.   In this presentation, I introduce the need of such methods and discuss case studies on protein dissociation and DNA base pair rolling, and show how the combination of these approaches can lead to a better understanding of biomolecular processes and biochemical signaling networks.

Collective cell dynamics play a crucial role in many developmental and physiological contexts. Recent works implicate the role of topological defects in tissue organization, but how collective dynamics and emergence of defects feedback on one another remains largely underexplored. In this talk, I will discuss how we utilize pancreas organoids to study rotational dynamics of epithelial spheres, which include persistent rotation, rotational axis drift and rotation arrest. Using a 3D vertex model, we demonstrate how the interplay between traction force and polarity alignment can account for these distinct rotational dynamics. Furthermore, our analysis shows that the spherical tissue rotates as an active solid and exhibits spontaneous chiral symmetry breaking. Using a continuum description, we demonstrate how the types and location of topological defects in the polarity field underlie this symmetry breaking process. Altogether, our work demonstrates how the subtle interplay between topological defects and collective cell dynamics govern tissue shapes and patterns, with implication for understanding and engineering tissue morphogenesis.

Throughout the tree of life populations have evolved the capacity to contend with sub-optimal conditions by engaging in dormancy, whereby individuals enter a reversible state of vanishing metabolic activity. The resulting "seed banks" serve as long-lived reservoirs of genotypic and phenotypic diversity. Of particular relevance is the case of microbial dormancy, which has a fundamental impact on the evolutionary, ecological and also pathogenic character of biological communities.

However, despite its ubiquity in nature, dormancy is a rather new paradigm in stochastic interacting particle systems. Here, it leads to novel effects, in particular based on the introduction of memory, resilience and diversity into the underlying systems. The resulting probabilistic structures, arising on different scales, are surprisingly rich, already when considering simple `toy models’, and lead to new universal behaviour.

In this talk, after providing some background on dormancy, I will present and discuss recent models for dormancy and seed banks in population genetics, evolution and ecology. Along the line, I will indicate topics for future mathematical and interdisciplinary research.

Systems biology is the combined use of experiment and theory with the goal of gaining a quantitative understanding of biological systems.
Accordingly, this presentation on systems biology work on bloodstream infections caused by human pathogenic fungi will be given jointly by an experimenter and a theorist. The focus will also be on the development of the research and its underlying philosophy, which should ultimately lead to the translation of basic research findings into clinical practice.

Embryonic development is arguably the most spectacular manifestation of the cells' ability to self-organize in spatial patterns while changing their identity to generate different organs. However, how cells generate these complex and robust dynamics is still largely unknown. 

In this talk, I will present our work aiming to address two key questions about mouse embryonic development: how do embryos decide where the head and the tail will form? How do they get rid of defective cells during their rapid growth? I will discuss how, to answer these questions, we analyze single-cell transcriptomic datasets and combine methods from machine learning and physics in close collaboration with experimental labs.

Today I will present our research on computational modelling of active filaments that dynamically reshape and cut cells across evolution. I will first discuss the physical mechanism of ESCRT-III filaments that drive cell division and perform cell trafficking in archaea and eukaryotes. I will then discuss the physical mechanism behind the dynamic self-organisation of active FtsZ filaments in the early stages of bacterial division. Througout, I will present comparisons of our models to live cell data. Beyond their biological context, our models can also help guide the design of artificial structures that are able to mimic life at the nanoscale.

Lipid transporters are integral membrane proteins and actively transport lipid molecules between the two leaflets of biological membranes. These proteins can move lipids between membranes via cavities that protect the lipids from the aqueous environment during transport. Several membrane protein families are involved in this process, comprising primary active transporters, namely ABC transporters and secondary active transporters from MATE and MFS families. Recently, we applied various computational tools, such as molecular docking and molecular dynamics simulations, to look at how these transporters recognize their substrates (1) and later how the binding process occurs (2,3). Using molecular dynamics simulations, we established different binding mechanisms that these transporters are employing to flip lipid and amphiphilic molecules depending on the substrates (2-4).

In this talk, I show how different pattern formation concepts may stand challenges arising from the current experimental research. Complexity of pattern formation in Hydra, which is a model organism of developmental biology, motivates analysis of different pattern formation mechanisms. Specifically, Turing-style morphogen-based models are compared to models coupling diffusive or mechanical signalling with nonlinear intracellular feedbacks.  The first class of models is based on reaction-diffusion-ODE systems, a coupling that yields destabilisation of all close-to-equilibrium (Turing) patterns and emergence of far-from-equilibrium patterns with jump-discontinuities. Patterning potential of mechano-chemical interactions is investigated using 4th order partial differential equations models coupling dynamics of diffusing molecular signals with tissue deformation. >span class="" lang="EN-GB"> >span class=""> model that explores patterning ability of Wnt/B-catenin-Dkk signalling system. The presented mutual inhibition model goes beyond the classical activator-inhibitor model and shows that a molecular mechanism based on mutual inhibition may replace the local activation/long-range inhibition loop. The new model is validated using a range of perturbation experiments. It resolves several contradictions between previous models and experimental data, and provides an explanation for the interplay between injury response and pattern formation.

Due to their role as barriers, epithelial tissues must be able to respond and adjust to various internal and external cues. As a result, epithelial cells exhibit rich active collective behaviours. Understanding these behaviours has been in the focus of active research in the physics of active matter. Over the past two decades, the vertex model has played a central role in modelling the mechanical behaviours of confluent epithelial tissues. In this talk, we address several possible scenarios of how activity can be introduced into the vertex model and how it affects its behaviour. We compare some of the findings with in-vivo imaging of the primitive streak formation in chick embryos. 

Mechanistic mathematical models are powerful tools in modern life sciences. Similar to experimental techniques, mechanistic models enable the evaluation of biological processes and hypothesis tests. Furthermore, mechanistic models allow the integrative evaluation of several data sets as well as the prediction of latent variables and future experiments. Yet, the scalability and need for large amounts of quantitative data has been encountered as a bottlenecks?
In this talk, I will outline how methods pioneered in the field of machine learning can assist with the development of large-scale mechanistic models. First, I provide an overview about the interfaces of mechanistic modelling and machine learning. Secondly, I will present tailor-made computational methods for training large-scale mechanistic models of cellular pathways using multi-omic data. This includes the application of mini-batch optimisation algorithms to scalable mechanistic modeling. Thirdly, I will outline methods for the use of low-grade data, e.g. information about qualitative properties, for the development of quantitative mechanistic models.

Normative theories and statistical inference provide complementary approaches for the study of biological systems. A normative theory postulates that organisms have adapted to efficiently solve essential tasks and proceeds to mathematically work out testable consequences of such optimality; parameters that maximize the hypothesized organismal function can be derived ab initio, without reference to experimental data. In contrast, statistical inference focuses on the efficient utilization of data to learn model parameters, without reference to any a priori notion of biological function. Traditionally, these two approaches were developed independently and applied separately. Here, we unify them in a coherent Bayesian framework that embeds a normative theory into a family of maximum-entropy ‘‘optimization priors.’’ This family defines a smooth interpolation between a data-rich inference regime and a data-limited prediction regime. Using simple examples from neuroscience and gene regulation, we demonstrate that our framework allows one to address fundamental challenges relating to inference in high-dimensional, biological problems.

Deep Neural Networks (DNNs) are state-of-the-art models for many vision tasks. We propose an approach to assess the relationship between visual tasks and their task-specific models. Our method uses Representation Similarity Analysis (RSA), which is commonly used to find a correlation between neuronal responses from brain data and models. With RSA we obtain a similarity score among tasks by computing correlations between models trained on different tasks. We demonstrate the effectiveness and efficiency of our method to generating task taxonomy on the Taskonomy dataset. Also, results on PASCAL VOC suggest that initializing the models trained on tasks with higher similarity score show higher transfer learning performance. Finally, we explore the power of DNNs trained on 2D, 3D, and semantic visual tasks as a tool to gain insights into functions of visual brain areas (early visual cortex (EVC), OPA and PPA). We find that EVC representation is more similar to early layers of all DNNs and deeper layers of 2D-task DNNs. OPA representation is more similar to deeper layers of 3D DNNs, whereas PPA representation to deeper layers of semantic DNNs. We extend our study to performing searchlight analysis using such task specific DNN representations to generate task-specificity maps of the visual cortex. Our findings suggest that DNNs trained on a diverse set of visual tasks can be used to gain insights into functions of the visual cortex. This method has the potential to be applied beyond visual brain areas.

Invasion processes play a role in various contexts. Successful invasion is often not guaranteed, but a matter of luck. Dr. Pokalyuk will talk on some biologically motivated invasion processes, corresponding asymptotic invasion probabilities and on strategies to prevent invasion.

Early embryogenesis is driven by complex spatio-temporal patterns that specify distinct cell identities according to their locations in the embryo. This process is remarkably reproducible, even though it results from regulatory interactions that are individually noisy.

Despite intense study, we still lack a comprehensive, biophysically realistic model for at least one biological system that could simultaneously reproduce quantitative data and rigorously explain the emergence of developmental precision. Moreover, traditional approaches fail to provide any insight as to why certain patterning mechanisms (and not others) evolved, and why they favor particular sets of parameter values.

We address both questions during early fly embryo development. Previous work has shown that the gap gene expression patterns in Drosophila optimally encode positional information. We therefore asked whether one can mathematically derive the gap gene network--without any fitting to data--by maximizing the encoded positional information. To this end, we extended our previous models into a generic, biophysically accurate spatial-stochastic model of gene expression dynamics, where gap genes respond to morphogen input signals and mutually interact in an arbitrary fashion, and optimized its parameters for positional information.

Firstly, our results show how the experimentally observed precision can be achieved with basic biochemical processes and within known resource and time constraints. Secondly, we show that a rich ensemble of optimal solutions exists and systematically analyse its characteristics, finding that some of the optimal solutions closely correspond to the real gap gene expression pattern. While numerous, optimal solutions still constitute a small subset of all possible solutions, and feature common properties. Finally, we explore a broad range of "mutated" optimal ensembles in which relevant components of the wild-type setting are altered or completely discarded, and systematically map out how this affects the encoded positional information and other relevant pattern properties; this allows us to rationalize the design of the wild-type gap gene system and the possible roles of its specific components. To our knowledge our work provides the first successful ab-initio derivation of a nontrivial biological network in a biophysically realistic setting. Our results suggest that even though real biological networks are hard to intuit, they may represent optimal solutions to optimization problems which evolution can find.

The delicate balance necessary for ensuring reliable specification of cell lineages is an intriguing problem in developmental biology. As an important paradigm in tissue development, the early mouse embryo cell fate decisions have been extensively researched, but the underlying mechanisms remain poorly understood. Current approaches to this problem still primarily rely on deterministic modeling techniques, although stochasticity is an inherent feature of this biological process.

As such, we are developing a multi-scale event-driven spatial-stochastic simulator for emerging-tissue development. We build up new simulation schemes for incorporating suitable tissue-scale phenomena, and we fix important parameters by using experimental values or numerical optimization to infer biophysically-feasible regimes. We first explore the characteristics of this system in a single-cell setting. We then extend the study to a multi-cellular setting in order to understand how positional information is robustly achieved and preserved. Our latest results indicate a potential signaling mechanism for reliable patterning emergence, despite strong constraints imposed by cell cycles. We are closely exploring how these signals redefine cell fates through the action of auto- and paracrine feedbacks.

For many applications in porous media and life science, problems are expressed as transient, non-linearly coupled systems of partial differential equations. Designing efficient solvers in this context is a delicate task, as suitable schemes for discretization, time integration and (non-) linear solution must be combined. We present a unified framework that combines linearly-implicit extrapolation with a scalable multigrid solver, which is thus suitable for HPC systems.

Efficacy and effectiveness of the approach are demonstrated for different applications from life sciences and for flow in the subsurface. We investigate robustness of the numerical methods, discuss suitable error estimators and analyze the scaling behavior.