In this talk, I will discuss a class of statistical methods for high-dimensional data that are termed kernel machines.   While they have been popularized in the machine learning and have found tremendous utility in various genomics contexts recently, in fact the mathematics that underlies the procedures date back to over 100 years ago.  In this talk, I will give a brief history of the development of kernel machines and show that one key property that arises is that of a metric. Given the availability of a metric for any particular data structure, a straightforward development of theory for testing for associations using kernel machines is available.  The methodology is fairly generic and can be applied to a wide variety of fields.   In this talk, we will describe applications of kernel machines to problems in multivariate genomic data fusion, metabolomics and neuroimaging genomics.  Time permitting, we will discuss an approach to causal inference using kernel machines.