2.2.6. Testing for significant differences in communities

Several analyses have been used to test for differences in microbial community composition (Schloss, 2008). It is likely, however, that new methods specifically developed for analysis of deeply sequenced microbial communities will complement or even replace the existing methods.

  1. Non-parametric MANOVA (NPMANOVA, (Anderson, 2001), also called Adonis in the vegan R package and QIIME) is a non-parametric analyses of variance that has been used to test for differences in microbial community composition.  In NPMANOVA and the following analyses the response variable is the dissimilarity or distance matrix calculated above and the independent variable is the group of samples being tested, (e.g. samples grouped by different regions of the bee gut or by an experimental treatment).  Multiple response permutation procedure (MRPP) is a related test for differences between communities.  NPMANOVA (Adonis) and MRPP are available in QIIME and require:
    a. a distance or dissimilarity matrix.
    b. amapping file.
    c. a categories list that details the category in the mapping file that will be tested (i.e. the independent variable).
  2. AMOVA (Excoffier et al., 1992) is another non-parametric analyses of variance that is available in mothur and tests whether the variation in each sample differs from the variation of the pooled samples.  To run AMOVA mothur requires:
    a a distance or dissimilarity matrix.
    b a design file that lists the sample name and corresponding group (i.e. the independent variable).

Although these tests are non-parametric, false positive results may be problematic and results should be interpreted with caution.