OTU picking is the clustering of the preprocessed reads into OTUs. The clusters are formed based on sequence identity. The identity threshold can be defined by the user. Sequences that are more than 97% identical are conventionally assumed to be derived from the same bacterial species/OTU. Other identity percentages can be used, depending on the granularity of the desired clusters and the known divergence in 16S sequences of the OTUs of interest. Three approaches for OTU picking exist. 1) de novo OTU picking groups sequences based on levels of pairwise sequence identity; 2) closed reference OTU picking aligns and groups sequences relative to a reference database, and sequences that are not >97% identical to a known reference are discarded 3) open-reference OTU picking starts with alignment to a reference database, but if the read does not match a known sequence it is not discarded but sent for de novo OTU picking. After the sequences have been clustered into OTUs and counted to estimate OTU abundance, a representative sequence is picked for each OTU. Each OTU is therefore represented by a single sequence and this will speed up downstream analysis. There are multiple choices to select a representative sequence. It can be the first sequence, the longest sequence, the seed sequence used in OTU picking, the most abundant sequence or a random sequence.
Software: UPARSE, QIIME
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.