RWNetSig

Abstract

Genetic disorders play an important role in cancer development. Especially, somatic variants, such as single nucleotide variations, insertions, deletions, and copy number variations, in specific genes, can promote tumorigenesis. Recently, NetSig has been developed that integrates existing protein interaction networks to predict cancer driver genes. However, NetSig is limited to use only directly interacting proteins. To address this issue, we developed a new statistic, called RWNetSig, that can be used to predict cancer driver genes by considering indirect as well as direct interactions in a network using the random walk with restart algorithm. In RWNetSig, indirectly interacting proteins are identified, the distance among them is calculated, the significance of their contribution to cancers is combined, and finally the RWNetSig score is calculated using the permutation statistical test. RWNetSig was evaluated and compared with Netsig using the Cosmic classic and the Cancer Gene Census datasets. We found that RWNetSig is superior to NetSig in identifying cancer driver genes. Our study reinforces the usefulness of network-based approaches in the field of the prediction of cancer driver genes. It will also contribute to gain new insight for deeper understanding of cancers.