Search

Ecological modelling can allow us to test hypotheses and answer questions in cases where empirical experiments may be impractical or impossible to perform. Larger models have relied heavily on the increasing power of computers over time to allow them to tackle larger problems. However, there is a hard limit to the speed at which a processor can operate and hence there is a limit to how performant a serialized simulation can be. To overcome this problem, programs can be written to utilize multiple cores to perform work in parallel. If a program is written in a highly parallelized way it may be adapted to execute on Graphical Processing Units. These have many cores and an extremely high compute throughput allowing for large scale simulations. By utilizing the power of GPUs, I have built several agent-based simulations to explore problems within ecology that have been intractable on regular hardware. Within this thesis, as well as explaining the advantages of GPU programming in an ecological modelling context I also explore three main problems in ecology. 1) The evolution of predictive adaptive responses and the mechanisms in which different informative cues may be integrated; 2) The evolution of dietary trade-offs, specifically the trade-off between fecundity and longevity utilizing the Geometric framework; 3)The evolution of trophic levels and possible causes of the trophic level ceiling, also using the geometric framework of nutrition. Although the problems have been approached before they have not had the compute behind them to simulate enough entities to produce emergent results. This is the first time GPUs have been applied to these problems and hence the first time realistic population sizes have been simulated providing unique insights.

Motivation Higher-order epistatic interactions can be the driver for complex genetic diseases. An exhaustive search is the most accurate method for identifying interactive SNPs. While there is a fast bitwise algorithm for pairwise exhaustive searching (BOOST), higher-order exhaustive searching has yet to be efficiently optimized. Results In this paper, we introduce BitEpi, a program to detect and visualize higher-order epistatic interactions using an exhaustive search. BitEpi introduces a novel bitwise algorithm that can perform higher-order analysis more quickly and is the first bitwise algorithm to search for 4-SNP interactions. Furthermore, BitEpi increases detection accuracy by using a novel entropy-based power analysis. BitEpi visualizes significant interactions in a publication-ready interactive graph. BitEpi is 56 times faster than MDR for 4-SNP searching and is up to 1.33 and 2.09 times more accurate than BOOST and MPI3SNP respectively. Availability Codes and data are publicly available on GitHub https://github.com/aehrc/BitEpi . BitEpi is also available on CodeOcean https://doi.org/10.24433/CO.3671084.v1 . Contact: Arash.Bayat@csiro.au - Denis.Bauer@csiro.au Supplementary information Supplementary data are available at bioRxiv online.

Presented are short descriptions of each parameter used in the model and their value(s).