Bohannan and Hughes 2003

Bohannan BJM, Hughes J. 2003. New approaches to analyzing microbial biodiversity data. Current Opinion in Microbiology. 6: 282-287.

These authors review the use of three broad approaches to studying microbial biodiversity in environments. The three are 1) parametric, 2) nonparametric, and 3) community phylogenetics. Each has advantages and disadvantages, and these authors suggest a combined approach may be most beneficial. Both 1) and 2) are based on Operational Taxonomic Units, to avoid the many problems of bacterial species identification, while 3) is based on molecular phylogenies, typically 16s rDNA.

Parametric approaches make simplifying assumptions and are based on some model of species richness in microbial communities; often this model is log-normal, in which some taxa are rare, some are abundant, and most are intermediate. These approaches extrapolate from patterns in a sample to the total environment. The obvious downside to parametric approaches is the vulnerability of the model to incorrect and difficult to test assumptions.

Nonparametric approaches avoid assuming any model, and instead are typically built on an approach analogous to mark-release-recapture. Sequences encountered more than once in a sample are recaptures, and the frequency of these doubletons is assumed to be related to how many unique sequences are present: more doubletons means fewer total sequences. As a downside, these approaches provide only a lower limit to actual richness, thus generally underestimating total diversity.

Community phylogenetics approaches avoid the OTU concept and thereby preserve useful data in the form of genetic information about sequences and sequence relationships. The downside of community phylogenetics approaches is they sample a clone library derived from the environment, not the environment directly, and can therefore not extrapolate from the sample to the environment.
This paper provides several useful examples of each approach, and supports the utility of Martin’s (2002) combined approach, which is what I would like to apply to my data to be collected in 2010. Figure 3 in this paper, for example, provides a useful overview of what Martin (2002) did, and how to make inferences about observed patterns.