RAMI: a tool for identification and characterization of phylogenetic clusters in microbial communities

Abstract

Motivation: The most common approach to estimate microbial diversity is based on the analysis of DNA sequences of specific target genes including ribosomal genes. Commonly, the sequences are grouped into operational taxonomic units based on genetic distance (sequence similarity) instead of genetic change (patristic distance). This method may fail to adequately identify clusters of evolutionary related sequences and it provides no information on the phylogenetic structure of the community. An ease-of-use web application for this purpose has been missing.

Results: We have developed RAMI, which clusters related nodes in a phylogenetic tree based on the patristic distance. RAMI also produces indices of cluster properties and other indices used in population and community studies on-the-fly.

Availability: RAMI is licensed under GNU GPL and can be run or downloaded from http://www.acgt.se/online.html.

Supplementary information: http://www.acgt.se/RAMI/SuppInfo.

Fig. 1.

Input and output flow of some relevant software for analysis of communities. Some necessary intermediate steps are not included as no output is generated. These include creating a patristic distance matrix which all tree reconstruction programs do, but not always output and the (local) alignment made by BLASTclust.

Fig. 2.

Explanation of indices produced by RAMI. (a) RAMI produces indices that describe properties within sequence clusters: X_{depth,nearest}, X_{depth,deepest} and X_distance. With a single sequence cluster, all indices will have a value of 0. (b) Analogous, the corresponding properties are measured between sequence clusters with the Y_{depth,nearest}, Y_{depth,deepest} and Y_distance indices. The value of Y_{depth,deepest} is dependent on the choice of outgroup. RAMI has the option to remove outgroup sequences from the analysis, which is important to get a proper value of the Y_distance index. Note that cluster 1 is paraphyletic.

Fig. 3.

Comparison of different clustering algorithms based on 269 full length γ-proteobacterial rRNA sequences from Greengenes manually curated core set (DeSantis et al., 2006). Identical cluster assemblies have a value of 0. In all pair-wise comparisons, distance thresholds for the respective algorithms have been set to produce the same number of clusters. Some of these are presented in the lower panel.

Fig. 4.

Geographic comparison of the NRI indices for eight bacterioplankton communities spread world wide. High positive values of this index indicate phylogenetic clustering, which was strongest in the Sargasso Sea sample. Negative values indicate overdispersion that was strongest in the Sydney sample. A two-tailed P-test showed that the Fiji, Plum Island Sound and Sargasso Sea samples were significantly structured at the P=0.05 level while the Sydney sample was significantly structured at the P=0.10 level.

Fig. 5.

Sequence clusters and geographical origins of sequences in marine bacterioplankton communities. Sequences in these plots are ordered according to the phylogenetic tree at the bottom. The order is the same in both horizontal and vertical directions. Sequence clusters are represented by squares along the diagonal and are surrounded by white space. Sequences in clusters are colored according to their geographical origin in such a way that when the origin of one sequence in the horizontal direction matches the origin of other sequences in the vertical direction (or vice versa), the area will be filled with the color representing the location. When all sequences in a cluster have the same geographical origin, the square will only have one color. When there are two or more geographical origins, the area where origins of sequences do not match are colored grey. The upper right part of the plot is a mirror image of the lower left part and is only provided for visualization purposes. (a) Verrucomicrobiae. A number of clusters are endemic to the Sargasso Sea (light blue), Fiji (green) and Chile (orange). (b) Cyanobacteria. The two largest clusters correspond to Synechococcus and the third largest to Prochlorococcus. A number of sequences with high rate of evolution are represented in the tree by long branches and in the plot with small squares that are separated inside the larger squares. For example, in the middle of the largest cluster, a number of rapidly evolving sequences have separated and are forming their own clusters. Thus, the largest cluster is paraphyletic. The RAMI indices for describing the clusters are presented in table 1.