A new theoretical angle to semi-supervised output kernel regression for protein-protein interaction network inference

Celine Brouard (1) , Florence d'Alché-Buc (1) , Marie Szafranski (2, 1)

1 IBISC - Informatique, Biologie Intégrative et Systèmes Complexes
2 ENSIIE - Ecole Nationale Supérieure d'Informatique pour l'Industrie et l'Entreprise

Protein-protein interaction network inference is addressed as an output kernel learning task through semi-supervised Output Kernel Regression. Working in the framework of RKHS theory for vector-valued functions, we establish a new representer theorem devoted to semi-supervised least square regression. We then apply it to get a new model and show its relevance using numerical experiments on artificial networks and a protein-protein interaction network dataset using a very low percentage of labeled proteins in a transductive setting.

Keywords: Protein-protein interaction; network inference; semi-supervised learning; kernel methods; operator-valued kernel

Domains: Machine Learning [cs.LG]; Bioinformatics [q-bio.QM]; Quantitative Methods [q-bio.QM]

Celine Brouard, Florence d'Alché-Buc, Marie Szafranski. A new theoretical angle to semi-supervised output kernel regression for protein-protein interaction network inference. International Workshop on Machine Learning in Systems Biology, Jul 2011, Vienne, Austria. ⟨hal-00832056⟩