BiRewire

BiRewire is an R package to enhance the analysis of combinatorial patterns within cancer genomic datasets, facilitating the identification of novel cancer driver networks. It builds on quantifying patterns like mutual exclusivity in gene mutations and evaluating the significance of these patterns using P-values computed under appropriate null models. The package improves upon traditional methods by introducing an innovative approximate lower bound for the number of switching steps required in the Monte Carlo method, which simulates datasets under a null model, preserving patient- and gene-wise mutation rates. This advancement, derived analytically, significantly reduces the computational expense previously dictated by an empirically deduced linear function of the total number of variants.

BiRewire utilizes a bipartite network representation of genomic datasets, applying Markov chain updates to alter network topology efficiently, thereby facilitating the generation of simulated datasets under the null model. This package demonstrates its utility and efficiency by applying to large real cancer genomics datasets, showcasing substantial time savings compared to existing methods while maintaining equivalent accuracy in P-value computations.