Fig. 2

Neighborhood queries enable recovery of relevant genomic content. a Left panel: recovery of each of three target genomes from podarV using queries at a variety of Jaccard distances from the target. Recovery is calculated as containment of target genome in query neighborhood. The solid lines represent logarithmic best-fit curves to the points. b Right panel: recovery of novel Proteiniclasticum content from podarV. Nucleotide k-mers from two of the three known P. ruminis genomes overlapped approximately a megabase of sequence in the query neighborhood, which also contained approximately 2.3 Mbp of unknown sequence; the third known genome, P. ruminis CGMCC, was omitted from the figure as it is 99.7% similar to P. ruminis DSM. Numbers are in thousands of k-mers, estimated via sourmash