RT Journal Article SR Electronic T1 Expanding discovery from cancer genomes by integrating protein network analyses with in vivo tumorigenesis assays JF bioRxiv FD Cold Spring Harbor Laboratory SP 151977 DO 10.1101/151977 A1 Heiko Horn A1 Michael S. Lawrence A1 Candace R. Chouinard A1 Yashaswi Shrestha A1 Jessica Xin Hu A1 Elizabeth Worstell A1 Emily Shea A1 Nina Ilic A1 Eejung Kim A1 Atanas Kamburov A1 Alireza Kashani A1 William C. Hahn A1 Joshua D. Campbell A1 Jesse S. Boehm A1 Gad Getz A1 Kasper Lage YR 2017 UL http://biorxiv.org/content/early/2017/06/19/151977.abstract AB Combining molecular network information with cancer genome data can complement gene-based statistical tests to identify likely new cancer genes. However, it is challenging to experimentally validate network-based approaches at scale and thus to determine their real predictive value. Here, we developed a robust network-based statistic (NetSig) to predict cancer genes and designed and implemented a large-scale and quantitative experimental framework to compare the in vivo tumorigenic potential of 23 NetSig candidates to 25 known oncogenes and 79 random genes. Our analysis shows that genes with a significantly mutated network induce tumors at rates comparable to known oncogenes and at an order of magnitude higher than random genes. Informed by our network-based statistical approach and tumorigenesis experiments we made a targeted reanalysis of nine candidate genes in 242 oncogene-negative lung adenocarcinomas and identified two new driver genes (AKT2 and TFDP2). Together, our combined computational and experimental analyses strongly support that network-based approaches can complement gene-based statistical tests in cancer gene discovery. We illustrate a general and scalable computational and experimental workflow that can contribute to explaining cancers with previously unknown driver events.