Application of database-independent approach to assess the quality of OTU picking methods
Assigning 16S rRNA gene sequences to operational taxonomic units (OTUs) allows microbial ecologists to overcome the inconsistencies and biases within bacterial taxonomy and provides a strategy for clustering similar sequences that do not have representatives in a reference database. I have applied the Matthew’s correlation coefficient to assess the ability of 15 reference-independent and -dependent clustering algorithms to assign sequences to OTUs. This metric quantifies the ability of an algorithm to reflect the relationships between sequences without the use of a reference and can be applied to any dataset or method. The most consistently robust method was the average neighbor algorithm; however, for some datasets other algorithms matched its performance.