Integrated Modeling and Optimization (Fundamental)

Course date
12-16 December 2022, Wageningen


  • The course will be held at: Forum Building (building nr 102), with number B0508.

Course coordinator

Natal van Riel, Eindhoven University of Technology
Dr. Dragan Bosnacki, Eindhoven University of Technology


Software: Matlab (with following toolboxes: Optimization toolbox, Statistics toolbox)

Course description

Living organisms are characterized by an amazing degree of hierarchical complexity. Although our ability to collect measurements at different spatial levels and time-scales has grown dramatically, it has become clear that only measurements cannot provide the answer to unravelling biological complexity. This is because the dynamical behavior of complex systems cannot be reduced to the linear sum of the functions of their parts. Hence, computational modelling is an absolute requisite to gain understanding of the mechanisms underlying patterns observed in experimental data, in particular when studying dynamic phenomena. Mathematical models allow in a relatively cheap way to generate and test hypotheses about these mechanisms. However, given the huge complexity and peculiar features of biological systems, it is necessary to carefully understand the specific modelling requirements they pose, in order to define what a good model should look like. In this way one could say that modelling is a craftsmanship, that can only be learned via intense exercising and ‘learning by doing’. In this course we offer the participants the possibility to learn and exercise the modeling process.

In validating models one always meet with the need to fit models to data. So, the parameters that are present in any realistic model have to be chosen based on comparison of model predictions with data. In this matching process optimization techniques are indispensable. That’s why a considerable part of this course is spent on getting you acquainted with the optimization techniques that are nowadays available and widely used. Numerical optimization also is the basis for so-called flux balance analysis (FBA), commonly used to study large metabolic networks. This type of models and their analysis and simulation is also introduced in the course.

The course is a mixture of theory sessions and computer practicals. The course is completed with an assignment to be finalized afterwards, for those who want to acquire 3 ECTS.


Daily schedule

9:30-12:30 lecture
12:30-14:00 lunch + meet-the-teacher
14:00-16:00 pen & paper exercises and/or computer practical
16:15-17:00 research talk + wrap up


Except Friday: 9.30-12.30 lecture & practical; 14.00 – 15.00 wrap up

On Wednesday there will be a dinner followed by a social event.


Day 1 (12 December) Introduction Systems Biology, math tools, optimization

Jaap Molenaar


  • brief intro participants at the start of the course
  • introduction of modelling with (ordinary) differential equations
  • introduction of optimization concepts and methods
  • practical (1): phase plane analysis
  • practical (2): Matlab intro (Hans Stigter, Dragan Bosnacki)

“Research talk” – Dragan Bosnacki discusses the topic linked to the assignment


Day 2 (13 december) Constraint-based modeling and analysis

Rob Smith


  • reaction kinetics
  • Flux Balance Analysis
  • linear programming optimisation
  • Dynamic FBA
  • practical using the COBRA package in MATLAB

Research talk: Sara Benito Vaquerizo (PhD candidate, WU) Modelling synthetic microbial communities for syngas fermentation to produce high-value chemicals


Day 3 (14 december) Networks, Network motifs, Network inference

Aalt-Jan van Dijk


  • network structure, motifs, dynamics
  • network inference
  • bias/variance: overfitting

Practical (Matlab): network inference with thickening – thinning

Research talk: Dr. Shauna O’Donovan (TUE) ‘Digital Twinning and Precision Nutrition’

On Wednesday 14 December there will be a dinner followed by a social event.


Day 4 (15 december) Numerical optimization, Parameter estimation, Sensitivity analysis

Hans Stigter


  • Maximum Likelihood Estimation (MLE)
  • linear and nonlinear least squares
  • optimization algorithms (local and global), steepest descent, Gauss-Newton, Levenberg-Marquardt
  • sensitivity (local)
  • Fisher Information Matrix
  • confidence intervals
  • experimental design
  • practical (Matlab)

Research talk: Dr. Dominique Joubert (WU) Local structural identifiability


Day 5 (16 december) Developing credible simulation models of biological systems

Natal van Riel


  • concepts on uncertainty and uncertainty quantification
  • profile likelihood analysis
  • bias-variance trade-off (regularization)
  • multi-start gradient search
  • experimental design


Introduction of the assignment (Dragan Bosnacki)


Learning objectives

The students will be provided with a theoretical basis, a variety of methods, and a computational hands-on experience to set-up systems biology models and handle numerical optimization.

In the course the students will learn:

  • To understand the common ground and the differences for applications of dynamic modeling in metabolic, regulatory, signaling, and multi-scale biological processes
  • How to set-up a dynamic model to represent biological networks using different interaction mechanisms
  • To implement, simulate and analyze dynamic network models
  • To understand the wide variety of problems in modelling that can be solved with optimization
  • To apply different types of numerical optimization methods
  • The combination of dynamic modeling and optimization to integrate experimental data in modelling, estimate model parameters and design experiments.
  • To understand how numerical optimization (linear programming) works in flux balance analysis to simulate metabolic network models.

Techniques include:

  • Nonlinear differential equations, numerical simulation, parameter sensitivity analysis.
  • Parameter estimation, identifiability, uncertainty quantification, experimental design, regularization.
  • Global and local search methods: steepest descent, Levenberg-Marquardt, genetic algorithms, linear programming.

Target audience

The course is aimed at PhD students with a background in bioinformatics, systems biology, computer science or a related field, and life sciences. Participants from the private sector are also welcome. A working knowledge of mathematics, especially differential equations, is recommendable, but we will distribute preparation material to be studied by students missing the required background. Furthermore, at the start we offer a math refresher to help those participants who are not (yet) involved in modelling on a daily basis.


Examples and computer practical make use of Matlab. A computer with a working version of Matlab is needed and some programming experience and knowledge of Matlab are required to take the course. A short introductory training in Matlab will be made available (online) for those without Matlab skills.


Registration for this course has been closed.