Prioritizing candidate genes by weighted network structure for the identification of disease marker genes

The use of microarray gene expression profiles for gene ranking is one of the most popular approaches to find marker genes associated with specific diseases. In addition, recently, other types of biological resources, such as gene annotations, bio-literature, and so forth, have been also explored along with the expression profiles. The GeneRank algorithm is one of such approaches that employs gene annotation data as well as expression scores to prioritize genes. Particularly, the GeneRank algorithm constructs an unweighted network structure from gene annotation data. Based on such network, it calculates ranking scores for individual genes according to their associated links and expression scores. In this work, our interest is to investigate the effectiveness of the weighted network structure generated from gene annotations for gene prioritization. For this purpose, we propose two novel weighting schemes to define the link strength between genes, called the Shared Functions (SF) link-weighting scheme and the Weighted Shared Functions (WSF) link-weighting scheme. The evaluation of the proposed schemes was done by applying them to prioritize candidate genes associated with prostate cancer. That is, from microarray expression profiles and gene annotation data, we produced ranking scores of individual genes based on the weighted network structure built by our proposed link-weighting schemes. As results, the top n-ranked genes were taken as our selection of marker genes associated with prostate cancer.