A parallel differential evolution algorithm for parameter estimation in dynamic models of biological systems

Abstract

Metaheuristics are gaining increased attention as efficient solvers for hard global optimization problems arising in bioinformatics and computational systems biology. Differential Evolution (DE) is one of the most popular algorithms in that class. However, the original algorithm requires many evaluations of the objective function, so its application to realistic computational systems biology problems, like those considering parameter estimation in dynamic models, results in excessive computation times. In this work we present a modified DE method which has been extended to exploit the structure of parameter estimation problems and which is able to run efficiently in parallel machines. In particular, we describe an asynchronous parallel implementation of DE which also incorporates three new search heuristics which exploit the structure of parameter estimation problems. The efficiency and robustness of the resulting method is illustrated with two types of benchmarks problems (i) black-box global optimization problems and (ii) calibration of systems biology dynamic models. The results show that the proposed algorithm achieves excellent results, not only in terms of quality of the solution, but also regarding speedup and scalability.