BiRewire

BiRewire performs bipartite network rewiring to generate null models that preserve patient- and gene-wise mutation rates for analysis of combinatorial mutation patterns in cancer genomics.

Key Features:

• Network Rewiring: Applies N consecutive switching-steps (Markov chain updates) to modify bipartite network topology while preserving patient- and gene-wise mutation rates.
• Efficient Computation: Uses an analytically derived approximate lower bound for the number of switching-steps to reduce computational cost compared to previous methods that scaled linearly with the total number of variants.
• Randomness Analysis: Provides functions to assess the randomness introduced across switching-steps to evaluate the stochastic properties of rewired networks.
• Performance Optimization: Implements efficient algorithms for the switching process to reduce time for P-value computations and other statistical analyses.

Scientific Applications:

• Null model simulation: Simulates datasets under null models that preserve mutation rates to evaluate the significance of combinatorial patterns, such as mutually exclusive gene mutations.
• Cancer driver discovery: Supports identification of novel cancer driver networks by enabling statistical assessment of mutation co-occurrence and exclusivity patterns in cancer genomics datasets.

Methodology:

Represent genomic datasets as bipartite networks and apply Markov chain switching-steps (N consecutive switching-steps) to rewire topology, using an analytically derived lower bound on the number of switching-steps and functions to analyze randomness, implemented with efficient switching algorithms to reduce time for P-value computations.