Specificity and sensitivity estimated with reference sequences. (A) The distribution of the bottleneck coverage versus the total nonsupport coverage among cycles inspected by DomCycle on a metagenomic data set simulated on 155 reference chromosomal sequences. The dashed line shows the threshold of significance (P < 0.01) for the global nucleotide test. Three cycles had global scores significantly above one and were selected as candidate dominant cycles. (B) The number of phantom cycles reported on the simulated metagenomic data set, comparing the performance of DomCycle, metaplasmidSPAdes (denoted by mpSpades), Recycler, and SCAPP. Two candidate dominant cycles passed the local nucleotide test. (C) The length distribution for phantom cycles reported on the simulated metagenomic data set. The two phantom cycles reported by DomCycle are <1 kb. (D) Two hundred simulated data sets were generated from 100 reference plasmids and 100 reference phages. Shown are recall estimates for DomCycle, metaplasmidSPAdes, Recycler, and SCAPP. Recall was defined as the number of runs with one or more correctly reported cycles divided by the total number of runs. (E) Precision estimates for DomCycle, metaplasmidSPAdes, Recycler, and SCAPP for the same data sets as in D. Precision was defined as the number of correctly reported cycles divided by the total number of reported cycles. (F) The number of reported cycles per sample for DomCycle, metaplasmidSPAdes, Recycler, and SCAPP when tested on reference plasmids and phages. DomCycle reported, at most, a single cycle in all cases. (G) Breakdown of the elements reported on the CAMI Low data set. (H) Recall and precision on the CAMI Low data set.
