We will be hosting mothur and R workshops throughout 2018. Learn more.
The get.communitytype command employs probabilistic modelling to cluster microbial communities into metacommunities, or enterotypes. The approach attempts to model the data as an increasing number of enterotypes using several techniques to identify the best fit to the data. The default method is based on Dirichlet multinomial mixtures as described by Holmes et al.. This approach was developed as it allows for clustering from unevenly sampled populations (i.e. different numbers of sequences per group). Several statistical techniques are employed to evalulate the fit of each number of metacommunities to the data, including the Bayesian information crition, Akaike information criterion and Laplace approximation. The Laplace approximation is recommended for selecting the best number of metacommunities.
Output files on the get.communitytype command include a relabund file describing the OTU distribution for each number of metacommunities, and a posterior file that displays the membership of each sample group to the metacommunities described. The results of the best fitting number of metacommunities is also summarised in a design file. It is important to note that mothur will find the optimal number of metacommunities by increasing the number of partitions until a local minimum Laplace value is observed. As a result a number of output files will be generated and it is up to the user to select the correct relabund file to describe the metacommunities.
The shared file is required.
mothur > get.communitytype(shared=test.0.03.subsample.shared)
The minpartitions parameter is used to .... Default=5.
The maxpartitions parameter is used to set the maximum number of partitions to be tested. If a minimum Laplace value is observed before reaching the maxpartitions value the command will exit before reaching it. Default=10.
The optimizegap parameter is used to set how many additional partitions to be tested after observing a new best fit. Default=3.
The groups parameter allows you to specify which of the groups in your shared file you would like analyzed. Group names are separated by dashes.
mothur > get.communitytype(shared=test.0.03.subsample.shared, groups=F3D0-F3D1-F3D11-F3D125-F3D13)
The label parameter is used to analyze specific labels in your input. Labels are separated by dashes.
mothur > get.communitytype(shared=test.0.03.subsample.shared, label=0.03)
The method parameter allows to select the method you would like to use. Options are dmm, kmeans and pam. Default=dmm.
The calc parameter allows to select the calculator you would like to use to calculate the distance matrix used by the pam and kmeans methods. By default the rjsd calculator is used.
The iters parameter allows you to choose the number of times you would like to run the subsample while calculating the distance matrix for the pam and kmeans methods.
The subsample parameter allows you to enter the size pergroup of the sample or you can set subsample=T and mothur will use the size of your smallest group while calculating the distance matrix for the pam and kmeans methods.
- 1.31.0 - First Introduced
- 1.32.0 - Name changed from get.metacommunity to get.communitytype.
- 1.32.0 - Bug Fix: order of sample names was not correct in .mix.posterior file if the shared file was not sorted.
- 1.33.0 - added dmm, pam and kmeans methods.
- 1.33.0 - added the method, calc, iters and subsample parameters.