In this tutorial we will continue with an OTU-based approach, for the phylotype and phylogenic approaches, please refer to the mothur wiki page.

In 16S metagenomics approaches, OTUs are clusters of similar sequence variants of the 16S rDNA marker gene sequence. Each of these clusters is intended to represent a taxonomic unit of a bacteria species or genus depending on the sequence similarity threshold. Typically, OTU cluster are defined by a 97% identity threshold of the 16S gene sequence variants at species level. 98% or 99% identity is suggested for strain separation.

Cluster sequences into OTUs

We will use the Cluster.split tool to perform clustering of the sequences into OTUs. With this approach, the sequences are split into bins, and then clustered with each bin. Taxonomic information is used to guide this process. The Schloss lab have published results showing that if you split at the level of Order or Family, and cluster to a 0.03 cutoff, you’ll get just as good of clustering as you would with the “traditional” approach. In addition, this approach is less computationally expensive and can be parallelized, which is especially advantageous when you have large datasets.

We’ll now use the Cluster tool, with taxlevel set to 4, requesting that clustering be done at the Order level.

Before we continue, let’s remind ourselves what we set out to do. Our original question was about the stability of the microbiome and whether we could observe any change in community structure between the early and late samples.

Because some of our sample may contain more sequences than others, it is generally a good idea to normalize the dataset by subsampling.

Note: since subsampling is a stochastic process, your results from any tools using this subsampled data will deviate from the ones presented here.