Facing the challenges of multiscale modelling of bacterial and fungal pathogen-host interactions

The authors present a review of the current approaches used to model host-pathogen interactions (HPI) and come of the challanges.  The paper begins with an overview of the benefits of multiscale models.  They authors argue that using multiscale models provide a deeper insighting the the interplay between hosts and pathogens while providing a mechanistic understanding of the interaction network. The importance of experimental studies are also highlighted in the paper.  The authors point that in order for multiscale models to work, there must be measurement data.  The data can be used to help formulate the model and also parameterize it. It also can be used to verification of the model at each scale is important and some attention should be drawn to gather data from the correct spatio-temporal scales. This should allow the model to make reliable and useful results.

A considerable amount of the paper discusses the approaches used to model multiscale HPI.  The first covered are game-theoretical models. These models include multiple scales but ignore time.  Typically game-theoretical models are used to investigate the outcome of interactions.  They assume that “players” are real-world entitites that take part in a “game” for which the goal is to optimize a pay-off.  These players choose a strategy and the optimal solution is discovered through the costs and benefits to the other player. This has been applied to evolving organisms (evolutionary game theory).  In HPI, evolutionary game theory have been used to study pathogen evaion strategies and host defense over time. Another approach that ignore time is constraint-based modeling. These models describe a biological system by a set of knowledge constraints which characterize possible behaviors.

The authors then consider models that integrate continuous time. These models are ones that we are familiar with ordinary differential equations (ODEs).  In addition to ODEs, partial differential equations also consider continuous time but also space.  However, if the model requires discrete time intervals, agent-based, state-based, cellular automata based Boolean or probabilistic models are useful.  These models take into account individual characteristics.  While they discuss each of these approaches individually, they also point out that combined approaches can help with modeling HPI.

The authors close with some of the challenges of using multiscale models in HPI.  One challenge is the successful combination of experimental and theoretical approaches.  An iterative cycle should be applied (see image above). Second, they point out that it is not the goal to completely mirror the real system in the model.  But rather, the biological question should dictate which parts are included and left out of the model. Next, the problem of different time scales and the number of features in each scale changes. Also, there are a large number of parameters is key to multiscale models. Last, there are ethical reasons why experiments cannot calibrate in human and animal systems.

Schleicher, J., T. Conrad, M. Gustafsson, G. Cedersund, R. Guthke, and J. Linde. 2016. Facing the challenges of multiscale modelling of bacterial and fungal pathogen-host interactions. Briefings in functional genomics:elv064-.