14 February 2013, 16:15
Ernst-Abbe-Platz 2, seminar room 3423

Kernel methods for small molecules and other structured data

Assoc. Prof. Juho Rousu (Aalto University, Helsinki Institute for Information Technology HIIT, “Kernel Machines, Pattern Analysis, and Computational Biology” Group, Finland)

Kernel methods, via the success of Support Vector Machines (SVM) are part of the standard toolkit in bioinformatics. Besides classification, they can be used for several data analysis tasks from regression, cluster analysis to predicting structured outputs. In this talk, I will give a tour of kernel-based data analysis with applications to bioinformatics. I will specifically concentrate in graph-based data arising in small molecule applications. Examples from our recent work such as enzyme function prediction, drug bioactivity prediction, and metabolite identification will be used to illustrate the approaches.