An essay I wrote for my Honours College Academic Writing course
Honours College: English for Academic Writing (HCPL204)
27 May 2020
How Simulation of Life Can Offer Insight Into the Nature of Nature
Insights into the molecular functioning of life have given rise to a view of the living world as merely a certain arrangement of molecules, out of which the properties of life emerge. Over recent decades, our understanding of molecular biology has made big leaps due to increased use of simulation of biological systems, slowly unwrapping the unfathomable complexity of life. However, the applications of biological simulation do not end at fundamental research. For instance, researchers claim they could simulate cancer cells with sufficient fidelity to decide in mere seconds what treatment of the tumor will be most effective in the next decade1. Simulation of life requires capturing the reactive nature of a biological system within a system of programmatic functions2, which could in essence be described as life, in silico. The advancements and questions in ‘executable biology’3, the field concerning the computational simulation of biological systems, give interesting insights into the nature of life itself and pose questions regarding our very reality. This paper will analyse the limitations and some emerging assumptions of the simulation of biological systems, and will subsequently examine the implications of the boundaries that divide biological simulation and the living systems it represents.
Firstly, the most obvious limitation of computational modelling is computing power. For the execution of a simulation a computer is required, which, as a consequence of limited natural resources and the physical rules that govern our universe, will always be limited in its computing power. Thus, the maximum number of calculations and operations that can be executed over a given time span is limited by physical factors4. However, that fact is not unique to the modelled system. Indeed, it is also the case in our natural world, where the matter surrounding us is also bound by fixed parameters: the laws of physics. Besides physical limitation, there is also the problem of our understanding of the systems we attempt to describe. The collective body of knowledge we have gathered throughout history is not, and might never be, complete. Therefore, a simulation of any natural system is also limited by the extent we understand that system. Furthermore, simulation of a biological system will always be less complex than its subject, considering the complexity of a simulation does not scale linearly, but exponentially. This escalating complexity can be attributed to the combinatorial nature5 of interactions within a biological system. Any element in a system interacts with any other element, illustrating how the addition of one more element does not result in a linear increase of complexity, but an exponential growth of the number of possibilities. Because computational resources and our understanding are limited, and the complexity of life endless, it can be argued that the making of assumptions about the behaviour of a system is inescapable. Secondly, the aforementioned limitations imply that assumptions and concessions must be made about the behaviour of the simulation. One assumption is that simplification and approximation are required. Thus, within reason, it can be assumed that a model can be simplified while still retaining sufficient predictive power, because increasing precision will yield diminishing improvements in the accuracy of the outcome. A second assumption could be that simulated systems are deterministic. A computed system will reach the same outcome when presented with the same input parameters. It must be noted, however, that a simulation is not required to be deterministic to be an accurate model for a natural system. On the other hand, it has been argued that no computed system is truly non-deterministic, as long as each and every one of the parameters is known, whereas others suggest that we can never know the true state of the universe the former argument seems to require6. This question surrounding the deterministic nature of simulations exemplifies that a system will represent the factors involved in its creation. Thirdly, whereas nature is the result of natural selection, computational models are consciously designed. The modelling of an emergent system—a phenomenon that cannot simply be viewed as a sum of its effects—is done by human hands, which might carry biases and inaccuracies into the system. As previously established, the limitations of simulations require concessions, which will always be based on incomplete information. Therefore, the creator and the simulation cannot be viewed as separate. In a sense, there is a poetic quality to this property of simulations—the simulation is not exclusively a representation of its subject, but also describes its author. The limits and preconditions surrounding a simulation of a natural system impose inevitable assumptions on the model.
Thirdly, when the limitations and questions surrounding assumptions are taken into account, it becomes clear that despite how complex an executable biological system becomes, we may still be able to reason about its limits and behaviour. One example of such reasoning about a program itself, is ‘static analysis’. Static program analysis within computer science entails the interpretation of a computed system without executing said system, which can also be applied to a program describing a biological simulation3. Therefore, the application of static analysis on biological simulations illustrates how the creation of computational models could, by itself, improve our understanding of the mechanisms that a simulation attempts to describe. The predictions made by a simulation can be considered not to be the only useful outcomes. Furthermore, it can be argued that the value of thinking about models and creating simulation need not merely be the prospect of ever achieving the perfect simulation. In addition, it is a fallacy to assume a simulated system is separate from the system it attempts to represent. Considering a phenomenon within our world is the subject of the simulation and the simulation is carried out in our world, total separation of the systems is non-existent, mirroring the relation between creator and creation. The physical system encloses the simulated system. Additionally, the fact that physical limitations enforce the simplification of simulations can illustrate that life itself is not much more than a machine that executes a biological program, yet life’s complexity is too great for any human to fully comprehend. Therefore, knowledge of the limitations of the systems we create ourselves can force us to reckon with the daunting complexity of life, and subsequently challenge us to create useful abstractions, that we can then apply to nature itself. Moreover, the limitations that govern computational simulation allow us to view our own world from a new vantage point.
In conclusion, the limitations that our physical world abides by are at least the same for any simulated biological system. These limitations and the systems they enclose can tell us about the boundaries of our simulations. Furthermore, they allow us to examine their behaviour, illustrating that the function of simulation is not only to predict outcomes, but also to allow us to better understand the mechanisms of its subject. Establishing the differences between a given simulation and its natural subject, teaches us a great deal about the validity of the simulation’s assumptions, andbut also lets us examine the assumptions that support nature itself. By examining the assumptions and mechanisms of the simulations we use to try to understand nature by prediction, we might concurrently learn more about the nature of nature.
Fisher, J., Harel, D., Henzinger, T. A. Biology as reactivity. Communications of the ACM, 54(10), 72-82 (2011). ↩︎
Kreuzaler, P., Clarke, M. A., Brown et al. Heterogeneity of Myc expression in breast cancer exposes pharmacological vulnerabilities revealed through executable mechanistic modeling. Proceedings of the National Academy of Sciences, 116(44), 22399-22408 (2019). ↩︎
Fisher, J., Henzinger, T. Executable cell biology. Nat Biotechnol 25, 1239–1249 (2007). ↩︎
Horvitz, E., & Zilberstein, S. Computational tradeoffs under bounded resources. Artificial Intelligence, 126(1), 1–4 (2001). ↩︎
Klamt, S., Stelling, J. Combinatorial Complexity of Pathway Analysis in Metabolic Networks. Mol Biol Rep 29, 233–236 (2002). ↩︎
Dürr, D., Goldstein, S., Zanghí, N. Quantum mechanics, randomness, and deterministic reality. Phys. Lett. A, 172(1–2), 6-12 (1992). ↩︎