Monday 28 September

13:30–14:00
Arrival, coffee and snacks
14:00–14:10
Welcome and introduction
14:10–14:35
Marko Radulović · Zagreb
Multiphysical effects on the fluid flow through a porous tube
14:35–15:00
Rebecca Crossley · Oxford
A likelihood-based framework for learning heteroscedastic noise and growth dynamics using biologically-informed neural networks
15:00–15:25
Filip Ivančić · Zagreb
Bacterial chemotaxis in biosuspensions with free surface: modeling and numerical strategy
15:25–15:55
Coffee break
15:55–16:20
Thomas Quinlan · Oxford
A Hybrid Cell-Based Model of Lipid-Immune Interactions in Atherosclerosis
16:20–16:45
Josip Tambača · Zagreb
Optimal design of elastic nanopore membranes for artificial kidneys
16:45–17:10
Pavle Goldstein · Zagreb
A Statistical Perspective on the Last Universal Common Ancestor (LUCA)
17:10 onwards
Discussion and networking

Tuesday 29 September

09:00–09:25
Arrival and coffee
09:25–09:50
Karla Bonic-Babic · Oxford
Estimating infectious disease outbreak risk under different superspreading mechanisms
09:50–10:15
Hrvoje Šikić · Zagreb
Superspreaders and the Markov Property
10:15–10:40
Robin Thompson · Oxford · online
Mathematical modelling at the end of an infectious disease outbreak
10:40–11:05
Lucija Krce and Marcela Mandarić · Split
BioFIDE project: From Bacterial Imaging to Topological Data Analysis
11:05–11:35
Coffee break
11:35–12:00
Abbie Evans · Oxford
Estimating how outbreak dynamics and intervention outcomes are shaped by different heterogeneity types
12:00–12:25
Janoš Terzić · Split
Contribution of gut bacteria to urinary bladder cancer
12:25 onwards
Lunch and discussion

Talks & abstracts

Marko Radulović

University of Zagreb

Multiphysical effects on the fluid flow through a porous tube

View abstract

In this talk, we study the steady-state laminar flow of an electrically conducting fluid under the influence of a magnetic field through a tube with permeable walls. The fluid flow is described by a non-linear system of equations given by the Navier–Stokes equations taking into account magnetic-induced effects. The main goal is to investigate the effects of the porosity of the tube and the magnetic field on the fluid flow. Using asymptotic analysis with respect to the small parameter representing the ratio between the tube’s thickness and its length, we construct an explicit first-order approximation for the axial and radial velocity of the fluid. The first-order approximation takes into account the porosity of the tube as well as the effects of the magnetic field on the fluid flow through the presence of the corresponding non-dimensional parameters, and we illustrate this by providing numerical examples. These presented results can be found in I. Pažanin, M. Radulović, Magnetic-induced effects on the fluid flow through a porous tube, Zeitschrift fur Naturforschung A 80 (2025), 1127–1136. I will also give a short introduction to future work on synovial fluid modelling.

Rebecca Crossley

University of Oxford

A likelihood-based framework for learning heteroscedastic noise and growth dynamics using biologically-informed neural networks

View abstract

In recent years, neural ordinary differential equation frameworks such as Biologically-Informed Neural Networks (BINNs) have shown promise for learning mechanistic laws from sparse data. However, most existing approaches implicitly assume homoscedastic Gaussian noise, and therefore do not account for potentially meaningful structure in biological variability. Here, we present an extension to the existing BINNs framework that includes a learnable noise model, allowing discovery of the noise model directly from data. Using population growth laws as an example, we demonstrate that the framework accurately recovers the underlying noise structure, scales uncertainty, and yields calibrated variance estimates that improve predictions of the underlying growth laws compared to existing approaches. As such, this work establishes a general likelihood-based framework for jointly learning dynamics and heteroscedastic noise within mechanistic neural network approaches.

Filip Ivančić

University of Zagreb

Bacterial chemotaxis in biosuspensions with free surface: modeling and numerical strategy

View abstract

Mathematical model governing bacterial chemotaxis in biological suspensions in bounded domains is extended and adapted to evolutionary domain. Domain with one boundary open to the atmosphere and free to move is considered. This is scenario of bacterial suspension in petri dish or in millimetric droplets. Design of numerical strategy preserving certain underlying structures of interest is proposed. Further extensions of the model will be discussed.

1. Ivančić, F., Sheu, T. WH, and Solovchuk, M.: The free surface effect on a chemotaxis diffusion-convection coupling system, Computer Methods in Applied Mechanics and Engineering 356 (2019): 387–406.

2. Hillesdon, A. J., and T. J. Pedley: Bioconvection in suspensions of oxytactic bacteria: linear theory, Journal of Fluid Mechanics 324 (1996): 223–259.

Thomas Quinlan

University of Oxford

A Hybrid Cell-Based Model of Lipid-Immune Interactions in Atherosclerosis

View abstract

Atherosclerosis is a disease driven by cholesterol accumulation within the arterial wall and involving coupled interactions between lipid transport and inflammation. Despite being a leading cause of cardiovascular disease, the spatial and stochastic mechanisms underlying plaque formation remain incompletely understood. To address this, we developed a hybrid continuum-discrete model of atherosclerotic plaque formation using the cell-based modelling framework Chaste. Our model incorporates feedback between lipid accumulation, cell death, and inflammatory chemokine production. We treat macrophages as discrete agents whose behaviour depends on the spatio-temporal distribution of lipoproteins, lipids, and the chemokine CCL2, whose dynamics are determined by a system of reaction-diffusion equations. The model is suitable for exploration of how cell-level processes influence local lipid and immune cell accumulation in early plaque development. Ongoing work focuses on exploring the model dynamics across parameter regimes.

Josip Tambača

University of Zagreb

Optimal design of elastic nanopore membranes for artificial kidneys

View abstract

Silicon nanopore membranes are a fundamental component of implantable bioartificial kidneys. While these membranes utilize micromachined slit-pores to achieve highly efficient blood filtration, the presence of these pores compromises their structural stiffness. In this study, we explore several strategies to mitigate this loss of structural integrity. These include optimizing pore dimensions, adjusting pore orientation to account for material anisotropy, and implementing optimal beam-modeled rib reinforcements. We formulate these design challenges as smooth parametric minimization problems targeting structural compliance. Finally, we prove the existence of solutions to these optimization problems and develop a gradient-based algorithm for numerical simulations.

Pavle Goldstein

University of Zagreb

A Statistical Perspective on the Last Universal Common Ancestor (LUCA)

View abstract

We introduce an analysis-of-variance statistical framework for analyzing multiple sequence alignments of protein families shared between two groups of organisms. When applied to present-day bacteria and archaea, the method yields a ranking score, estimating likelihood for their presence at time of LUCA population. By varying bacterial and archaeal subsamples, we assess the robustness of these inferences and gain some information on metabolic and other pathways potentially present in LUCA. Finally, our results provide a perspective on evolutionary processes associated with the archaeal–bacterial divergence.

Karla Bonic-Babic

University of Oxford

Estimating infectious disease outbreak risk under different superspreading mechanisms

View abstract

Superspreading can strongly affect whether an introduced infection fades out or causes a major outbreak. In this talk, I will compare two mechanisms of superspreading: superspreading events, where transmission varies through time, and superspreading individuals, where individuals differ in their transmission potential. I will show that, even when the two models have the same reproduction number, dispersion and offspring distribution, they can produce different estimates of the probability of a major outbreak after conditioning on the same early incidence data. I will then present an ensemble approach that combines these estimates according to the posterior support for each model, and illustrate how uncertainty in the mechanism of superspreading can be incorporated into real-time outbreak-risk estimation.

Hrvoje Šikić

University of Zagreb

Superspreaders and the Markov Property

View abstract

We present some consequences of the Markov property in modelling the superspreaders in infectious diseases models. We introduce the notion of an accidental superspreader, being purely the consequence of the specific behavior of the branching Markov process vs. the real superspreader. All authors contributed equally to the paper and are affiliated with the University of Zagreb, Faculty of Science, Department of Mathematics.

Robin Thompson

University of Oxford

Mathematical modelling at the end of an infectious disease outbreak

View abstract

Determining when an infectious disease outbreak can be declared over safely is a critical challenge for public health policy advisors. Although end-of-outbreak declarations permit the relaxation of control measures, premature declarations risk subsequent disease resurgence. As a result, there has been increasing interest in developing quantitative frameworks for real-time estimation of the probability that additional cases will occur. These approaches can be applied towards the end of an outbreak, following the apparent final reported case, to determine the risk associated with declaring the outbreak over.

I will present several mathematical modelling approaches for assessing the probability that additional cases will occur at the apparent end of an outbreak. First, I will describe a method based on stochastic compartmental models, in which repeated model simulation can be used to derive practical “rules of thumb” for the minimum waiting time without cases required before declaring an outbreak over. Second, I will present an exact probabilistic framework based on stochastic branching process models, in which disease incidence time series data are combined with contact tracing information to calculate the probability of additional cases. I will also show how this framework can be extended to settings in which contact tracing data are unavailable. Finally, I will discuss a renewal equation based approach for estimating the probability of additional cases directly from disease incidence time series data. This flexible framework was applied in the context of the 2017 Ebola outbreak in the Democratic Republic of the Congo to assess when interventions could be relaxed safely.

Together, the methods presented provide a rigorous quantitative basis for informing public health decision making at the end of an infectious disease outbreak.

Lucija Krce and Marcela Mandarić

University of Split

BioFIDE project: From Bacterial Imaging to Topological Data Analysis

View abstract

The BioFIDE project connects research on bioactive molecules and nanomaterials with mathematical analysis of biological systems. Here, we address a practical question: how can microscopy images be used to quantify bacterial responses to antibacterial treatment?

Lucija Krce, principal investigator of bioFIDE, will introduce the project and present efforts to develop a relatively simple procedure for preparing bacterial cells for scanning electron microscopy. Using conventional and alternative imaging modes, we compare untreated Escherichia coli and Staphylococcus aureus cells with cells exposed to an antibacterial surfactant belonging to the quaternary ammonium compounds—a class widely used in disinfectants.

Building on these experiments, Marcela Mandarić will present an approach that assesses cell shape and uses topological data analysis to examine bacterial surface texture. Persistent homology describes features of the bacterial surface that are not captured by cell outlines alone. Statistical tests assess differences between treated and untreated cells. The study illustrates how experimental imaging and mathematical methods can be combined to move beyond visual inspection towards a quantitative description of treatment-associated changes in bacterial morphology.

Abbie Evans

University of Oxford

Estimating how outbreak dynamics and intervention outcomes are shaped by different heterogeneity types

Mathematical epidemiology. Title and abstract coming soon.

Janoš Terzić

University of Split

Contribution of gut bacteria to urinary bladder cancer

View abstract

Exposure to environmental pollutants and the composition of the human microbiota may influence the risk of cancer. Like drugs, pollutants are metabolized in the body, which can alter their carcinogenic effects and where they reach in tissues. Although gut microorganisms can transform many chemicals and affect their metabolism, their role in pollutant-induced cancer remains unclear. We showed that reducing gut bacteria changes the metabolism of cigarette smoke-associated nitrosamines and substantially decreases the development and severity of nitrosamine-induced bladder cancer in mice. Using bacterial isolates, human gut microbial communities, and mouse models, we identified specific bacteria that contribute to this carcinogen metabolism. We also found that this activity varies between individuals and applies to several related nitrosamine carcinogens. These findings suggest that gut microbiota can influence how environmental chemicals contribute to cancer development. Understanding these interactions could help improve cancer risk assessment and identify new microbiome-based approaches for cancer prevention.