Figure 5
From: A systematic search for discriminating sites in the 16S ribosomal RNA gene
Details of phylum classification. Panel A shows how the number of mis-classifications drops as more and more sites are selected by the site selection algorithm. The error levels out at around 100 mis-classifications, and 50 selected sites seems to be enough to achieve this error rate. Panels B-F are PLS-plots of the sequence data, and the various panels show the same data from different perspectives. In panel B we plot the data in a coordinate system spanned by PLS-component 1 and 2, in panel C it is spanned by component 3 and 4 and so on. Each dot corresponds to a sequence, and the colors represent the true class label for each sequence, indicated by the legend in panel B. This figure is based on the Greengenes data, but the RDP and SILVA data gave similar results.