One widely used qualitative method to estimate coverage is a rarefaction curve, sometimes also called a collector or complexity curve. This method relies on the observation that the curve of rarefied counts of any feature (for example, operational taxonomic units, named species, predicted genes, functional categories or even short motifs) should plateau if the sample is close to saturation. Use of rarefaction curves in microbial community studies was popularized by tools such as mothur (Schloss et al., 2009) and recently extended to include accurate projections at higher sequencing efforts by preseq (Daley and Smith, 2013), which allows the estimation of coverage across features (arithmetic mean). However, this technique and others like it typically rely on a high-quality assembly, comprehensive reference data sets or both, which are often unavailable for complex or poorly characterized communities (with the probable exception of ribosomal RNA (rRNA) genes). Moreover, the preseq projection is optimized for single-species data sets and, therefore, does not scale for mixtures of species, making it insufficient for accurate estimations with complex metagenomic data sets. Without accurate projections, rarefaction curves can only be used to determine whether a data set is close to saturation, a useful but insufficient assessment of coverage. Performing this task with rRNA genes is also problematic; largely because their high sequence conservation frequently masks important levels of genetic and ecological differentiation among closely related organisms (Caro-Quintero and Konstantinidis, 2012).

Another approach is to estimate the coverage of one or a few target species in the metagenomic data set using simple statistical approaches such as the Lander–Waterman expressions (Lander and Waterman, 1988), while ignoring the remaining genomes of the community. Such methods are useful in studies targeting specific species in a community in order, for instance, to recover complete genomes. The main drawbacks of this approach include a lack of implemented software and the requirement of reliable estimates for genome size and abundance of the target species, which often poorly represent the community as a whole. No matter how limiting this approach may appear, it can be applied to many available metagenomic data sets, is based on robust statistical frameworks (Wendl et al., 2012) and the interpretation of coverage is straightforward: breadth of the genome covered by sequencing reads.

Log in or create a free account to read this content

Gain free access to this article, as well as selected content from this journal and more on nature.com

Access through your institution

or

Sign in or create an account
Continue with Google
Continue with ORCiD