A biologist’s perspective on modelling

Before I came out to Colorado State University, the first thing that jumped to mind when I heard about modelling in science was ‘predicting the future’. Predicting the future – that would be fascinating wouldn’t it? That’s what the Intergovernmental Panel on Climate Change tries to do with Earth using mathematical modelling of the climate system. However mathematical modelling can do more than surmise about the future (which is difficult to do, even with a good model and good data).

I’ve learned that modelling can help us to understand nature in a way that may be inaccessible to experimentation (e.g. take too many experiments to reach the same conclusion), and it can reveal gaps in our knowledge. Modelling can do this because, being virtual, we can run thousands or millions of ‘experiments’ in mere minutes, an enormous task outside of a computer. In biology, we can vary the information that goes into the model to see what happens. For example, if we were interested in what would happen if we made photosynthesis less sensitive to temperature in a plant, we could simply change the numbers that describe the temperature response of photosynthesis and run a simulation. An experiment with real plants would test the limits of genetic engineering, not to mention take countless hours to run.

Now suppose instead that we were confident that we knew everything about photosynthesis and how a plant uses photosynthesis to grow – how would we know if we were right? Well, if a model could accurately describe the results of an experiment with many different environmental conditions, it would suggest that we were correct. However, what if the model just couldn’t quite match the results of a single group in our experiment, but could perfectly describe everything else? This indicates that we are missing something that is fundamentally different in the underlying biology. In this case we can try to figure out what could be different, and test our ideas using the model, or we could use the information that the model gives us to run another experiment.

Modelling then becomes part of the toolkit of science, rather than a goal, that we can use to further our knowledge. As Donald Rumsfeld (the origin of the phrasing is murky) once said, “there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don’t know we know”. As I’ve come to realize thus far in my trip, this message describes exactly how we can use mathematical modelling – we construct a model of the known knowns (e.g. the current state of knowledge on photosynthetic biology), when the results of the model are off by a consistent quantity, this tells us the known unknowns (e.g. we know that there is something we don’t understand that determines how photosynthesis contributes to growth), and lastly, if model predictions are off in only a few special cases, then that reveals the unknown unknowns to us. Modelling thus allows us to figure out what we didn’t know we didn’t know so that we can experiment and uncover new knowledge.

And so, if you have ever doubted or misunderstood the use of modelling in science, keep in mind that it can be a powerful tool to further experimentation in a given field and our understanding of science as a whole.

Stay safe and stay informed,
Joe