Disease gene prioritization through semi-supervised transductive methods

We propose novel methods to rank genes with respect to Cancer Modules (CMs), that is sets of genes coordinately under or overexpressed in cancer diseases obtained from the Molecular Signatures Database (Re and Valentini 2011).

We used both protein-protein and domain-domain interaction networks enforced through the predictions of a classifier, and functional interaction networks constructed with comparative genomics techniques to rank genes with respect to CMs, by applying classical random walk techniques (Re and Valentini, 2012), label propagation algorithms and kernelized score functions (Re and Valentini, 2012). In particular kernelized score functions compare favorably to state-of-the-art semi-supervised machine learning methods, both in terms of average AUC and precision at a fixed recall (Re and Valentini, 2012).