Published online by Cambridge University Press: 05 April 2013
Computational and statistical tools play an important role in materiomics, to provide insights in the underlying processes that allow certain materials to outperform other materials. In this chapter, we discuss numerous methods that allow the analysis of materiomics data. Specifically, we describe the use of statistical tests, ranking and data mining approaches, model learning and testing, as well as experimental design and the exchange of experimental results. Also, we review some of the important publications in this field from the past 15 years, organizing them according to the type of material descriptors that were used.
Basic principles of data analysis
Computational methods play an ever more important role in the study of material function. Partly, this is due to the increased scale of the experiments being performed, with an accompanying need for automated analyses. But the move from low-throughput towards high-throughput experiments entails more than just testing more materials simultaneously. The extra information these experiments produce is slowly catalysing a transition to a more rational approach to material discovery, in which not just material screening plays a role but also material modelling. Materials and their environments are approached as systems that can be modelled and thus explored in silico. This ‘systems approach to material research’ has been termed materiomics. This transition is certainly needed given the size of the materiome that one wants to explore: many material parameters can be varied and combined into a practically infinite palette of combinations. This far surpasses even the reach of high-throughput screenings. The question that will be addressed in this chapter is: how can we efficiently make use of our capability to perform high-throughput experiments, to explore and characterize such a large search space?