Hypothesis Generation from Heterogeneous Datasets

The last decade has seen a rise in the rapid accumulation of large-scale data from both genomic technologies and from increased use of electronic health records. These advances have been accompanied by opportunities for automatic hypothesis generation in translational research; however, integrating and mining these highly heterogeneous datasets remains challenging. This chapter addresses the major principles and methods that are associated with providing effective solutions to a broad range of these problems. Indeed, these principles include issues of representation, biological scales of measurements, feature selection, and statistical approaches to address the curse of dimensionality, and approaches of integration that we divide into corroborative versus fusion approaches.