Microbiome

The GuMI platform is a microfluidic gut-on-chip device that simulates human gut-microbiome interactions under continuous fluid flow [1]. While this physical bio-chip generates rich data on how gut bacteria influence human gut cells, these experiments are not easily accessible. To address this issue, this project develops a digital twin of the GuMI platform. The digital twin, generating in silico predictions, will allow cheap exploration of yet untested experimental setups, allowing prescreening and refinement of empirical hypotheses.

Specifically, this project develops a PDE-based digital twin of the GuMI experimental setup. You will formulate and solve a spatiotemporal computational model incorporating coupled oxygen gradients, fluid dynamics, metabolite diffusion, microbial kinetics, and host tissue response within the microfluidic geometry [1]. By calibrating the model against experimental data, you will assess different experimental conditions and explore the complex crosstalk between the host and the gut microbiome.

Objectives

1. Data extraction & geometry setup: extract parameter values and map the device geometry.
2. PDE model development.
3. Model calibration & validation: fit numerical parameters to experimental data.
4. In silico exploration: simulate unmapped experimental regimes to predict system behaviour.

References

1. Zhang et al. (2021). https://doi.org/10.1016/j.medj.2020.07.001

Expectations

Good programming skills (Python).

Work environment

The student will join a diverse team of researchers from the Informatics Institute (IvI) and the MetaHealth project (NWO), including experts in AI development, modelling, experimental design, and validation. You will contribute directly to a computational model aimed at predicting spatiotemporal host–microbiome interactions and optimizing gut-on-chip experimental design.

Spatial organization within the gut microbiome shapes host–microbiome interaction, substrate utilization, and overall community stability. High-resolution protocols such as SAMPL-seq (Spatial Anchored Microfluidic Particle Sequencing) now capture microbial co-localization at ~40 µm resolution but remain observational [1] — static snapshots that cannot isolate the underlying physical drivers (reaction-diffusion gradients, advection, competition) or predict perturbation outcomes.

This project explores how a mechanistic, spatially resolved model of the gut micro-environment could reproduce and extend beyond what SAMPL-seq data show, moving from descriptive co-occurrence patterns toward a predictive, causal understanding of spatial niche formation and collapse. Further exploration of the calibrated model will allow investigation of dietary effects or drug exposure on the reshaping of gut spatial ecology.

Objectives

1. Survey the state of the art in spatial gut microbiome modelling and high-resolution spatial profiling methods.
2. Model development and calibration.
3. Scenario simulation: explore how dietary or drug-exposure perturbations reshape gut spatial ecology.

References

1. Uppal et al. (2025). https://doi.org/10.1186/s40168-025-02279-4

Expectations

Good programming skills (Python).

Work environment

The student will join a diverse team of researchers from the Informatics Institute (IvI) and the MetaHealth project (NWO), including experts in AI development, modelling, experimental design, and validation, contributing to a computational model aimed at unravelling gut spatial ecology.

The role of the oral microbiome in dental caries formation remains unclear. Recent in vitro work reveals a striking mismatch: introducing Streptococcus mutans drastically shifts microbiome community structure, yet overall biofilm production remains unchanged [1]. No mechanistic model currently explains this disconnect.

This project aims to build a mechanistic, predictive model using multi-species in vitro datasets. By modelling microbial growth kinetics, metabolic interactions, and biochemical reactions, you will move beyond descriptive statistics to uncover the causal drivers of microbiome resilience. Finally, you will simulate in silico intervention scenarios, such as varying sugar exposure frequencies, to predict the key mechanisms involved in this dysbiosis.

Objectives

1. Data curation: preprocess and analyse in vitro datasets.
2. Model construction: translate biological mechanisms into a computational dynamic system.
3. Parameterization: fit model parameters to map the specific role and interactions of S. mutans.
4. Scenario simulation: test dietary interventions and targeted perturbations in silico.

References

1. Mazurel et al. (in preparation).

Expectations

Good programming skills (Python).

Work environment

The student will join a diverse team of researchers from the Informatics Institute (IvI), ACTA (Prof. Dongmei Deng) and the MetaHealth project (NWO). You will contribute to a computational model aimed at unravelling the role of the oral microbiome and be involved in the prevention of dental caries.

The human microbiome is a complex ecosystem where thousands of microbial species interact, exchange metabolites, and influence host physiology. Unravelling these cross-feeding networks and host–microbe dynamics is vital for understanding disease mechanisms and designing targeted therapies.

However, translating massive, fragmented biochemical databases into fast, executable simulation models remains a significant software engineering bottleneck. This project aims to design and implement an automated software pipeline and knowledge graph framework — the "Microbiome Map" — that integrates reaction networks and solute cross-feeding data to power high-speed, spatiotemporal simulations of microbial communities.

Objectives

1. Construct a unified knowledge graph pipeline: automate data integration that extracts and links metabolic reactions, species traits, and nutrient exchange capabilities from curated resources such as AGORA2 into an extensible graph database.
2. Automate simulation model extraction: build software modules and query interfaces that dynamically translate subgraphs of microbial interactions into structured models ready for constraint-based and individual-based simulation platforms.
3. Accelerate spatiotemporal simulations: design scalable pipelines using parallel computing, optimized spatial solvers, or machine learning surrogates (such as graph neural networks) to drastically improve multi-species simulation performance.

References

1. Heinken, A., et al. "Genome-scale metabolic reconstruction of 7,302 human microorganisms for personalized medicine." Nature Biotechnology 41.9 (2023): 1320–1331.
2. Bauer, E., et al. "BacArena: Individual-based metabolic modeling of heterogeneous microbes in complex communities." PLoS Computational Biology 13.5 (2017): e1005544.
3. Dukovski, I., et al. "A metabolic modeling platform for the computation of microbial ecosystems in time and space (COMETS)." Nature Protocols 16.11 (2021): 5030–5082.

Work environment

The student will join an interdisciplinary team bridging software engineering, artificial intelligence, and computational biology. You will work on software architecture, graph engineering, and high-performance computing, with the chance to build open-source tools that directly accelerate biological discovery and healthcare modelling.

Genome-scale metabolic models can predict bacterial growth, uptake, and secretion through flux balance analysis, optimizing an objective function. This function mostly tries to maximize the biomass produced by the bacterium [1]. The assumption is usually examined under well-mixed conditions, whereas bacteria grow inside nutrient and product gradients they generate themselves, where the choice of objective may matter considerably more.

This project develops a spatial single-species growth model that couples reaction-diffusion transport of nutrients and secreted metabolites to genome-scale metabolism solved locally — that is, dynamic flux balance analysis on a grid using a Cellular Potts Model [2]. You will implement and verify the coupling, then compare objective functions and determine whether they leave distinguishable spatial signatures, such as colony expansion rate, front morphology, nutrient penetration depth, and where metabolites are secreted.

Objectives

1. Model development: formulate and implement the coupled PDE and constraint-based solver for one species.
2. Numerical verification: mass conservation, mesh convergence, and convergence in the coupling timestep.
3. Objective comparison: simulate maximal biomass, maximal ATP, parsimonious FBA, and a rate–yield trade-off.
4. In silico exploration: identify spatial observables that discriminate between objectives and could be measured experimentally.

References

1. Orth et al. (2010). https://doi.org/10.1038/nbt.1614
2. Sheraton et al. (2018). https://doi.org/10.1128/aac.02544-17

Expectations

Good programming skills (Python) and familiarity with numerical methods for PDEs. Constraint-based modelling can be learned during the project.

Work environment

The student will join a diverse team from the Informatics Institute (IvI) and the MetaHealth project (NWO), spanning mechanistic modelling, metabolic reconstruction, and experimental validation. You inherit a working hybrid PDE and cell-model codebase rather than starting from scratch, and receive an account on the Snellius national HPC cluster for parameter sweeps. The model forms the foundation for multi-species spatial simulations under development in the group.

The mouth and the gut host distinct microbial communities, yet oral bacteria are transmitted to and colonize the large intestine extensively in healthy people: of 310 species tested across five countries, 125 were prevalent in both habitats [1]. Which species make this crossing, and which metabolic and phenotypic traits allow it, has never been catalogued systematically. Computational modellers of either community currently rebuild this information by hand from scattered, incompatible databases.

This project builds that resource and then interrogates it. You will scrape taxonomic, phenotypic and metabolic data from public microbiome and strain resources, hand-label a reference set of cross-niche taxa from the primary literature, and train models that predict cross-niche presence from computable traits. The curated dataset feeds directly into the oral and gut microbiome models under development in the group.

Objectives

1. Data extraction & reconciliation: scrape oral and gut resources, harmonize taxonomy across sources, define defensible presence criteria.
2. Gold standard curation: hand-label a reference set of cross-niche taxa against the primary literature.
3. Predictive modelling: predict cross-niche presence from metabolic and phenotypic traits, cross-validated against a simple baseline.
4. Interpretation & release: identify which traits carry the signal and release a documented, openly available dataset.

References

1. Schmidt et al. (2019). https://doi.org/10.7554/eLife.42693.001

Expectations

Good programming skills (Python) and comfort with web scraping and data cleaning. No prior biology is required.

Work environment

The student will join a diverse team from the Informatics Institute (IvI) and the MetaHealth project (NWO), spanning data science, mechanistic modelling and experimental validation. The work is openly released and contributes to a shared reference layer for ongoing oral and gut microbiome modelling in the group.