The large amounts of biological information generated using advanced high-throughput experimental techniques continue to motivate the design of suitable methods for valuable knowledge mining. Finding proper means to examine and analyze such information allows better understanding of normal biological processes as well as uncovering malfunctions that trigger various diseases. Several computational approaches were developed to complement the experimental work which is often restricted by high time and cost requirements. In this paper, we consider the problem of disease- gene association and we propose a methodology based on a classification approach which integrates protein-protein interaction network topology features and biological information collected from various data sources. When applied on a dataset of multiple disease types and using the Naive Bayes classifier, our method achieves an AUC score of 0.941. We also consider two case studies of type II diabetes mellitus and breast cancer. The experimental results greatly favor our approach.