Fig. 1.

Overview of the LazyB assembly pipeline. (a) Short Illumina reads are filtered to represent only near unique k-mers and subsequently assembled into unambiguous unitigs. Long Nanopore reads (ONT) can be optionally scrubbed to include only regions consistent to at least one other read. For larger data sets scrubbing can be handled on subsets efficiently. Mapping unitigs against Nanopore reads yields unique “anchors” between them (b). An undirected graph (c) is created by adding Nanopore reads as nodes and edges between all pairs of reads sharing an “anchor”. Each edge is assigned a relative orientation, depending on whether the “anchor” maps in the same direction on both Nanopore reads. Cycles with a contradiction in orientation have to be removed before choosing a node at random and directing the graph based on its orientation. As Nanopore reads that are fully contained within another do not yield additional data, they can be collapsed. Contigs are extracted as maximally supported paths for each connected component (d). Support in this context is defined by the number of consistent overlaps transitive to each edge. Final contigs (e) can be optionally polished using established tools