The Virtual Laboratory
Like all biological systems, the microbial world is composed of many levels, all of which are intricately interconnected. Subtle changes in the microscopic level may blow up to shape the entire ecosystem, and vice versa, animal behaviour may determine which microorganisms and mobile genetic elements thrive. A major challenge in microbial biology is unpacking all these levels and trying to understand how changes in one level propagate to another. But as you can imagine, this is all very complex, and we may be looking for diamonds in the dark. Which strategies can we use to shine light on these complex systems?
Whether experimentally or theoretically, the process of ”shining light” can either be very precise, or very broad. In other words, we are either:
…using a flashlight: Designing an elegant experiment or model to study an otherwise well-understood system.
Often with a hypothesis as to how the system may respond, i.e. the light is directed
Or… using a torchlight: Experimentally manipulating a complex system with many unknowns
Often without a preconceived notion of what will happen to the system, i.e. the light is undirected
In reality, no experiment or model is precisely at either extreme. However, whereas experimental biology appears to occupy all possible positions along this axis, theoretical biology appears to be predominantly laser-focussed. That is to say: by making an interesting simplification of reality, can we prove our hypothesis works if nothing else interferes. For example, by designing simple models of mutualists and parasites, we may find possible steady states of their interactions.
Again, if nothing else interferes.
I want to argue that there is one important flaw with the above strategy, which is that apes like ourselves are not very good at estimating which variables may or may not interfere, or how complex interactions pan out. In fact, the answer to many questions in biology is often not very intuitive, or only intuitive in hindsight. So that leaves us with the second strategy: exploring a dark cave with a torch light.
As said, exploratory strategies are quite common in experimental biology, but much less so in theoretical biology. After all, if the number of parameters in the model gets out of hand, we may be better off actually getting our hands dirty on the real system. I however want to argue that, when trying to unpack a complex system with many unknowns, complex multi-level models are a surprisingly productive starting point.
Going bottom-up
When trying to model a complex system, the first question in one’s mind may be: where on Earth do I start? The problem is: if there are so many potential unknown interactions between different levels and components, how can we model them?
The answer is: going bottom-up. For example, instead of defining which type of cells exist and how they interact, we can instead model the nature of this interaction itself. Cells may take up resources, and convert them into other types of metabolites. Some of these metabolites may be used to make DNA, proteins, and cellular membranes, while others may simply be waste products. Such waste products may or may not be used by another cell, or could even be used to harm other cells when cleverly produced in large amounts.
In the above example, we do not define with which “rates” cells inhibit one another’s growth, but instead define a set of metabolic rules with sufficient degrees of freedom. Then, we can simply let the system live its life and the system will tell us a story. Or perhaps it will tell us that we forgot something. Either way, this is a very educational journey! And once such a system is in place, you have your own “Virtual Laboratory” with which you can experiment!
By building multi-level biological systems from the bottom up, you have your own “Virtual Laboratory” with which you can experiment!
Evolution can solve the “parameter curse”
Even in the relatively simple case described above, the number of parameters quickly explodes. In my personal experience, simply designing a random cell with a random metabolic network will, more often than not, result in the system collapsing. With this many degrees of freedom (and thus, many parameters!), how do we know what value they should all have?
Well, maybe we don’t need to know which parameters work and which do not. Remember, evolution tuned nature’s parameters perfectly fine on its own. As evolution is a very simple process to emulate (random mutation and selection), we can simply leverage that power to tell us which parameters work, and which don’t. This can serve as aproof of principle; there exists a universe where evolution shapes the system in this way. Of course, biological relevance should then also be investigated thoroughly.
A treasure chest of unexpected behaviour
Even though your system may be very complex, you probably have some intuitions about what your system can and can’t do. Generally, we can therefore split our results into two subcategories:
Results and Results+
Does the system work (results) and show the expected self-organisation (results+)? If not, why not?Results++
Does the system have emergent properties that were unforeseen? Are they biologically reasonable and interesting?
The result++ in the image above refers to the “self-mixing” effect of cross-feeding behaviour observed in Virtual Microbes.
The term “Results++” is coined by Paulien Hogeweg in her course Computational Biology. The first plus refers to the emergent properties of the system that one may anticipate, while the second plus refers to unexpected outcomes. Thus, similar to experimental systems, bottom-up models can be leveraged to generate new hypotheses. Of course, this concept is not only applicable to microbiology, but also applies to other fields such as the origin of life, animal behaviour, and development.
Conclusion
I argue that with the computational power available to us now, the Virtual Laboratory is an important tool that will allow us to validate our expectations, generating new hypotheses along the way.
