Efficient detection of generalized bibubbles in bidirected graphs

Develop an efficient algorithm to identify all generalized bibubbles in a bidirected gene graph—i.e., minimal pairs of oriented genes (x, y) whose enclosed vertex set satisfies the specified reachability symmetry and minimality conditions—such that the method scales to very large graphs with tens of millions of nodes, including minigraph-cactus and PGGB graphs.

Background

The paper introduces a rigorous definition of generalized bibubbles to capture local gene order, copy-number, and orientation variations directly on bidirected gene graphs. Existing superbubble-finding algorithms apply to directed graphs and do not generalize cleanly to bidirected graphs, especially in the presence of inversions.

The authors present practical methods based on net graphs and cycle equivalence and implement a working approach that performs well for human-scale gene graphs (~20,000 genes). However, they note that this implementation will not scale to much larger graphs produced by whole-genome pangenome tools like minigraph-cactus or PGGB, leaving scalability as an unresolved challenge.