Kernel-based methods are powerful methods for integrating heterogeneous types of data. mixKernel provides methods to combine multiple kernels for unsupervised exploratory analysis or to select features in a kernel for unsupervised analysis or kernel-output prediction. For the multiple kernel problem, different solutions have been implemented to obtain a consensus kernel or a kernel that best preserves the original topology of the data. mixKernel also integrates kernel PCA to visualize similarities between samples in a nonlinear space and from the multiple source point of view. Functions to assess and display important variables using permutations are also provided in the package.
mixKernel is available on CRAN and information are provided on its website. It is fully compatible with the mixOmics package and a tutorial describing the approach can be found on mixOmics web site. Its source code is available on ForgeMIA