Role of Centrality in Network-Based Prioritization of Disease Genes

High-throughput molecular interaction data have been used effectively to prioritize candidate genes that are linked to a disease, based on the notion that the products of genes associated with similar diseases are likely to interact with each other heavily in a network of protein-protein interactions (PPIs). An important challenge for these applications, however, is the incomplete and noisy nature of PPI data. Random walk and network propagation based methods alleviate these problems to a certain extent, by considering indirect interactions and multiplicity of paths. However, as we demonstrate in this paper, such methods are likely to favor highly connected genes, making prioritization sensitive to the skewed degree distribution of PPI networks, as well as ascertainment bias in available interaction and disease association data. Here, we propose several statistical correction schemes that aim to account for the degree distribution of known disease and candidate genes. We show that, while the proposed schemes are very effective in detecting loosely connected disease genes that are missed by existing approaches, this improvement might come at the price of more false negatives for highly connected genes. Motivated by these results, we develop uniform prioritization methods that effectively integrate existing methods with the proposed statistical correction schemes. Comprehensive experimental results on the Online Mendelian Inheritance in Man (OMIM) database show that the resulting hybrid schemes outperform existing methods in prioritizing candidate disease genes.