UC Santa Cruz

Sequence graphs have emerged as an important tool in two distinct areas of computational genomics: genome assembly and pangenomics. However, despite this shared basis, subtly different graph formalisms have hindered the flow of methodological advances from pangenomics into genome assembly. In genome assembly, edges typically indicate overlaps between sequences, with the overlapping sequence expressed redundantly on both nodes. In pangenomics, edges indicate adjacency between sequences with no overlap—often called blunt adjacencies. Algorithms and software developed for blunt sequence graphs often do not generalize to overlapped sequence graphs. This effectively silos pangenomics methods that could otherwise benefit genome assembly. In this paper, we attempt to dismantle this silo. We have developed an algorithm that transforms an overlapped sequence graph into a blunt sequence graph that preserves walks from the original graph. Moreover, the algorithm accomplishes this while also eliminating most of the redundant representation of sequence in the overlap graph. The algorithm is available as a software tool, GetBlunted, which uses little enough time and memory to virtually guarantee that it will not be a bottleneck in any genome assembly pipeline.