Scimm + PhyScimm :: Center for Bioinformatics and Computational Biology

Scimm is an unsupervised method for clustering metagenomic sequences using interpolated Markov models in a framework similar to the k-means clustering algorithm. Currently, we suggest LikelyBin and CompostBin to initially partition a subset of the sequences in order to initialize the IMMs, and the package provides wrapping code for these programs. PhyScimm initially partitions the sequences using a supervised classification algorithm Phymm, which performs better in high complexity datasets and when the species encountered are well-represented in public databases.