Optimisation and landscape analysis of computational biology models

Sodecl

“…Consequently, optimisation methods are often employed to find the particular parameter combinations that most closely reproduce the experimentally-measured behaviour of the system of interest [Paláncz et al 2016]. Performing this parameter optimisation step in a robust, systematic fashion is a critical step in the construction and analysis of biological models [Akman et al 2008[Akman et al , 2010[Akman et al , 2012Avramidis and Akman 2017;Doherty et al 2017;Lillacci and Khammash 2010] because determining the optimal parameter values enables alternative models to be ranked and experimentally testable predictions to be formulated [Ashyraliyev et al 2009; Avramidis and Akman 2017; Cedersund and Roll 2009;Cullen et al 1996;Johnson and Omland 2004;Lillacci and Khammash 2010;Slezak et al 2010;Sun et al 2012]. The assumptions made in constructing a given model can then be rigorously assessed, and insights obtained into how the model could be modified so as to improve the accuracy of its predictions [Ashyraliyev et al 2009; Avramidis and Akman 2017; Cedersund and Roll 2009;Lillacci and Khammash 2010;Slezak et al 2010].…”

“…The computational cost can also be significantly increased by the particular parameter optimisation method employed; for example, when using a population-based method-such as an evolutionary algorithm (EA) or a particle swarm optimiser (PSO)-the equations have to be integrated multiple times over different parameter values to explore the underlying fitness landscape [Cedersund et al 2016;He and Yao 2001;Witt 2008]. Additionally, a resampling approach might be required to mitigate the effects of noise and uncertainty in model evaluation, thereby increasing the computational load further [Doherty et al 2017;Fieldsend 2015].…”

Data science for the scientific life cycle

“…As an example he explained how he had used this approach to find the optimal parameters for a mathematical model for the circadian clock (Aitken and Akman, 2013). Akman also described how it is possible to search for possible gene regulatory networks that can explain existing experimental data (Doherty, 2017), and then select new experiments to help distinguish between these alternative hypotheses (Sverchkov and Craven, 2017). For instance, the algorithm might suggest performing a certain gene knockout experiment, followed by RNA-seq, to gain more information about the network structure.…”

Reducing the complexity of mathematical models for the plant circadian clock by distributed delays

“…Despite a large collection of advanced molecular experiments (Flis et al, 2015), many biochemical details -such as the values of chemical reaction constants -are still unknown. Parameter optimisation is therefore indispensable in reproducing quantitative features of the experimental data (Locke et al, 2005a;Pokhilko et al, 2010;Akman et al, 2008Akman et al, , 2010Akman et al, , 2012Foo et al, 2016;De Caluwé et al, 2016;Doherty et al, 2017). The optimisation procedure, however, rapidly becomes computationally intractable as the network architecture gets more complex, due to the concomitant increase in the number of unknown parameters.…”