We are pleased to announce that the Oncodrive methods family has a new member: OncodriveROLE, an approach to classify cancer drivers into loss of function and activating roles. OncodriveROLE joins and complements the previously developed methods that identify cancer driver genes from the list of somatic mutations in cohort of tumors (OncodriveCLUST and OncodriveFM).
Cancer driver genes come in two main flavors: those that contain driver alterations which cause the loss of function (LoF) of the gene product (for instance, in tumor suppressor genes like TP53 or CDKN2A), and those with driver alterations that increase or change the activity or function of the protein product, such as oncogenes like PIK3CA or BRAF.
Distinguishing between these two classes of driver genes is very important to understand tumorigenesis in patients and has profound implications for therapeutic decision making and for the development of targeted drugs.
For these reasons some time ago we started the development of a method that could accurately classify cancer driver genes in those with activating mutations (Act) and those with Loss of Function mutations (LoF). We first assessed the capability of 30 gene features related to the pattern of genomic alterations across tumors to distinguish between Act and LoF cancer genes. Among those with higher discriminatory power we found features that measure the proportion of truncating mutations in the gene, the ratio of missense mutations within mutation clusters and the ratio of CNA gains versus losses affecting the gene. We then combined highly discriminatory features to train a random forest classifier, that we named OncodriveROLE. The method shows an accuracy of 0.93 and Matthew’s correlation coefficient of 0.84 classifying genes in the Cancer Gene Census. We plan to incorporate OncodriveROLE into the IntOGen-mutations pipeline; meanwhile the method is available at http://bg.upf.edu/oncodrive-role/. On the website you may obtain the classifier for your own use and also browse the results of the two cancer driver sets presented in the article in detail.
We are happy to announce that the manuscript describing OncodriveROLE is now published in Bioinformatics and freely available online. I will present the method in a talk at the ECCB meeting in Strasbourg next moth. I’ll be happy to discuss with any one interested on the method during the meeting and answer as well via the comments in the blog.