Shedding light on the black box models of the cell

ion as realistic as possible. Such models are extremely useful and can help to identify new properties or elements of biological systems. What is considered the first mathematical model in biology falls into a different category. In the year 1202, long before calculus was invented, the mathematician Fibonacci tried to predict how many rabbits could breed over time in ideal circumstances starting from a pair of female–male rabbits. He assumed that the animals would never die and that females could only give birth to a new pair of female and male rabbits each time. This is clearly an oversimplified model that is far from biological reality. It renders a series of numbers – the celebrated ‘Fibonacci numbers’ – that are very different from the numbers of animals that would have been determined empirically. The discrepancy between prediction and observation in this case highlights the difficulty of building models with predictive power prior to having appropriate empirical evidence. Engineers often use models to explore the ‘design space’ of new products before actually building them, but are generally working with systems assembled from well-defined components. In the context of Synthetic Biology, which has among its goals the standardized design and construction of novel dynamic systems, such as genetic circuits, the underlying components (‘parts’) are often quite poorly characterized, and this leads to a corresponding decrease in the predictive power of our models. Synthetic biologists have managed to address the problem of the discrepancies between model predictions and experimental observations by following the so-called engineering cycle. This consists of employing successive iterations of design, modelling, construction and testing of novel biological systems. Although this iterative cycle has proven to be a useful tool, it is time and resource intensive. Thanks to advances in DNA synthesis and screening, it is now possible to automate the design and construction of certain genetic circuits (Nielsen et al., 2016). This method shows great promise and allows significant reductions in the time required to obtain working constructions. It requires, however, the use of well-characterized parts that may not be suitable for all applications. Besides automation, another possibility to facilitate the design of genetic circuits is to improve the way we build models. The question that remains open and that many people are trying to answer is why are models inaccurate? I have posed this question to different audiences showing for illustration the side-by-side comparison of the theoretical and experimental behaviour of a synthetic oscillator. The model predicts elegant and periodical oscillations, whereas the results obtained in the laboratory show minimal damped oscillations that are Received 25 November, 2016; accepted 25 November, 2016. *For correspondence. E-mail j.jimenez@surrey.ac.uk; Tel. +44 1483 68 4557. Microbial Biotechnology (2017) 10(1), 43–45 doi:10.1111/1751-7915.12489 a 2017 The Authors. Microbial Biotechnology published by John Wiley & Sons Ltd and Society for Applied Microbiology. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. bs_bs_banner