Our team has been dedicated to systematically constructing cancer driver prediction methods using biostatistics and artificial intelligence technologies. In recent years, we have developed a series of published algorithms and software, including CanDriS, CN/CS-calculator and MODIG, to predict cancer drivers at mutation, gene, and network levels. Here, we provide web tools for these methods for user convenience, allowing users to freely explore cancer drivers in their mutation data through a standardized workflow.
CanDriS is a method using an empirical Bayesian procedure to calculate the posterior probability of a site being cancer-driving and profiling the potential cancer-driving sites for thousands of tumor samples from TCGA and ICGC.
CN/CS-calculator contains two CN/CS models (CN/CS-Pos, CN/CS-H) which can be used to comprehensively analyze the selection pressures of different sites in the cancer genome. Specifically, CN/CS-Pos is a fusion model based on mutation frequency and selection pressure for driver gene prediction, which exploits the complementary advantages of methods based on different biological hypotheses. CN/CS-H can detect genes with weak positive selection for some amino acid sites despite the presence of lethal mutations. These genes were considered as mini-driver genes. According to the relationship between the selection pressure of genes and the functions they played in tumor progression, CN/CS-calculator can be applied to identify driver genes and mini-driver genes.
