Optimization in computational systems biology via high performance computing techniques

Abstract

The aim of computational systems biology is to generate new knowledge and nnderstanding about complex biological systems by combining experimental data with mathematical modeling and advanced computational techniques. The development of dynlIDlic modeJs (also known as reverse engineering) is one of the current key issues in this area.. In recent years, research has been focused on scaling-up these kinetic models. In this context, the problem of parameter estimation (model calibration) remains a very challenging task. The complexity of the underlying models requires the use of efficient solvers to achieve adequate results in reasonable computation times. Global optimization methods are use<! to solve these types of problems. In particular, metaheuristics have emerged as an efficient way of solving these hard global optimization problems. However, in most realistic applications, metaheuristies still require a large computation time to obtain ""ceptable results. This Thesis presents the design, implementation and evaiuation of novel parallel metaheuristics with the focus on parameter estimation problems in computational systems biology. In particular, we propose new cooperative metaheuristics based on the well known Differential Evolution and Scatter Search algorithms. The design of the novel approaches aim to achieve a proper balance between exploration (global search) and exploitation (local search) abilities. We show how the cooperation between parallel searches improves the behavior of the individual optiInizers, improving the quality of the obtained solutions while decreasing the time-to-solutiou. We also explore adaptive strategies in order to iucrease the robustness of the algorithms. We preseut encouraging results for the proposed metaheuristics considering very challengiug large-scale benchmark problems. Botb traditioual high performance computing (HPC) parallel and distributed architeetures aud new cloud infrastructures have been used to evaluate the proposals.