Unsupervised Two-Way Clustering of Metagenomic Sequences

Figure 2

Illustration of a two-way Poisson mixture model for metagenomic data. Each cluster represents a species and is modeled as a distribution of words comprising it. Each word follows a different distribution. However, not all words in a class have significantly different parameters. Therefore, the words can be divided into groups and words within the same group can be constrained to have identical parameters.