Skip to content

Systems biology, digital twins and AI

Welcome to the course "Systems biology, digital twins and AI". In this course we will explore how mathematical modelling can be used to create digital twins and how these twins currently are being used in research. The course will give you a broad understanding of the underlying theory needed to develop such digital twins, and introduce AI solutions used to strengthen the abilities of the twins. The course will introduce you to various modelling formats and help you familiarise yourself with the strengths and limitations of the different types of models. Nonlinear properties of dynamical systems will be introduced to raise awareness of the importance to understand the dynamical system you are modelling. Further, various tools and strategies will be introduced and evaluated, including optimization, parameter estimation and uncertainty estimation. Multi-level modelling will introduce the ability of modelling to cover different scales - from cell level to whole-body level - which is essential in creating a digital twin that describes how the whole biological body operates. These modelling techniques are connected to the fields of bioinformatics, AI, and machine learning and introduces Hybrid modelling, an essential solution to handle the enormous availability of information in physiology while still keeping the underlying mechanistic understanding of the system. Lastly, we introduce the use of digital twins in basic research, drug development, as a replacement for animal experiments, and in clinical and everyday applications.

Course layout

The course is designed in three segments: an information gathering phase, a phase to apply the new knowledge, and lastly a dugga to test your knowledge.

The first part starts with an introductory week. After that, the main four weeks of the course consists of the lectures, scheduled interactive classes, and the computer exercises. Before each interactive class, you need to prepare by watching the pre-recorded lectures outside of the scheduled time. The material from the recorded lecture is then discussed on the interactive classes. Apart from this, you also have a practical part where you apply the theory by doing the computer exercises. Note that you are expected to spend more time on the computer exercises than what is scheduled. You produce a written report for the computer exercises, answering some questions from each topic of the computer exercises. Finally, each week will end with discussions around the concepts from both the lecture and the computex exercise.

After the four main weeks, you will do an individual assignment where you delve a bit deeper in a subject of your choice. You are free to choose any subject within the course, but you can of course find inspiration from the lectures, interactive classes, and computer exercises. In the end, you will have a seminar where you present your chosen topic and take part in discussion on both your own and other projects.

Lastly, there will be a dugga that test your knowledge on the course contents.

Course content

The overall purpose of the course is to advance the understanding of methods and approaches within data-driven mechanistic systems biology and apply this understanding to real world problems within biomedicine. The course contains the topics listed below:

  • Different modelling formats: ordinary differential equations, differential algebraic equations, Boolean models, nonlinear mixed-effects modelling, stochastic models.
  • Properties of nonlinear dynamic systems: numerical methods for simulation, continuation, and bifurcation analysis; basin-of-attraction, transients, and stationary behaviors.
  • Basic concepts and methods within data-driven mechanistic modelling: formation of cost functions, optimization, statistical tests.
  • Methods of uncertainty analysis: profile-likelihood, core predictions, Markov Chain Monte Carlo.
  • Parameter identifiability analysis: difference between structural and practical identifiability.
  • Multi-level models: flat modelling, modular modelling.
  • Hybrid modelling: basics within bioinformatics, AI, and machine learning to calculate risks and generate perturbed images and responses; combinations of statistical and mechanistic models.
  • Usage areas of modelling and digital twins: in basic research, for drug development, for replacement of animal experiments, and for clinical and everyday applications within eHealth.