A discriminative approach to structured biological data

This paper introduces the first author’s PhD project which has just got out of its initial stage. Biological sequence data is, on the one hand, highly structured. On the other hand there are large amounts of unlabelled data. Thus we combine probabilistic graphical models and semi-supervised learning. The former to handle structured data and the latter to deal with unlabelled data. We apply our models to genotypephenotype modelling problems. In particular we predict the set of Single Nucleotide Polymorphisms which underlie a specific phenotypical trait.