Figure 5
From: Clustering metagenomic sequences with interpolated Markov models
FAMeS accuracy. Cluster accuracy statistics for FAMeS Arachne-assembled contigs. We ran SCIMM at a range of values for k, the number of clusters, resulting in consistently high accuracy on the low complexity simLC dataset and variable accuracy on the medium complexity simMC for reasons discussed in the text. We ran PHY SCIMM, ignoring IMMs matching genomes in the mix at the strain and species levels (PHY SCIMM-strain and PHY SCIMM-species above). When some IMMs from the same species can be expected, accuracy is far greater, demonstrating that clustering with supervised help relies on models for similar genomes.