This thesis addresses the problem of finding robust, fast and precise learning methods for noisy, incomplete high-dimensional biological data by means of so-called kernel methods. Kernel methods are at the heart of many ...