3. Pooling and arranging • bubbler

Pooling taxonomy

library(bubbler)
library(ggplot2)

The taxonomic level is selected with the taxa_level argument, when generating the relative abundance table. By default it is set to “Phylum”. Let’s see how many different taxa are present in a table.

library(dplyr)
rel_abund <- rel_abund_qiime(counts_q, taxa_q, metadata_q, taxa_level = "Genus")

rel_abund %>%
    summarise(unique_taxa = n_distinct(taxon)) # see how many unique taxa there are 
#> # A tibble: 1 × 1
#>   unique_taxa
#>         <int>
#> 1         196

There are 196 unique taxa in the relative abundance table. If we attempt to plot this right away, the legend will include all these taxa. The table must be filtered to pool taxa below a given threshold. The pool_taxa function can be given a numerical threshold or the number of taxa that should be displayed. In the latter case, the threshold is determined automatically.

rel_abund %>%
    pool_taxa(n_taxa = 12) %>% # threshold set to 0.0038
    bar_plot(position = "fill")

Pooling, by default will remove any meta data that was attached to the relative abundance table. Specify keep_metadata to keep it. Setting label to FALSE will set the threshold label in the legend to “Other”.

rel_abund %>% 
    pool_taxa(n_taxa = 8,
              keep_metadata = TRUE,
              label = FALSE) %>%
    bar_plot(position = "fill", x_var = "body_site")

Arranging

Taxa

In the legend, taxa are ordered alphabetically, which is the default for ggplot2 when dealing with character variables. Sorting the taxa by their abundance would be more meaningful. Using the arrange_taxa function, taxa can be arranged in ascending order, or flipped by setting the order argument to “top” or “bottom”. The threshold order can also be changed by setting the pooled argument to “top” or “bottom”. After ordering, it becomes clear that Bacteroides is the most abundant taxon in the gut, and Streptococcus among the other samples.

rel_abund %>% 
    pool_taxa(n_taxa = 8,
              keep_metadata = TRUE) %>%
    arrange_taxa(pooled = "top", order = "bottom") %>% # arrange by abundance
    bar_plot(position = "fill", x_var = "body_site") +  
    labs(y = "mean_rel_abund")

Variable

Another option is to arrange the samples by a variable, which can be useful when the data includes metadata. Observe the similarities between the above and below plots. The above plot represents the mean relative abundances among the levels of body_site, and the bottom plot represents the relative abundance by sample, ordered by body_site.

rel_abund %>% 
    pool_taxa(n_taxa = 8,
              keep_metadata = TRUE) %>%
    arrange_taxa(pooled = "top", order = "bottom") %>% # arrange by abundance
    arrange_var(var = "sample_id", levels = "body_site") %>%
    bar_plot(position = "fill")

Samples by taxa abundance

Samples can be arranged by taxa abundance. I first saw this technique used by Chenxin Li, in their github repo: Friends Dont Let Friends Make Bad Graphs. Combined with bubbler::arrange_taxa, bubbler::arrange_sample_by_taxa will partition samples by their most abundant taxa. In the absence of meta data, this is a good preliminary visualization to produce.

rel_abund %>% 
    pool_taxa(n_taxa = 16) %>%
    arrange_taxa() %>% 
    arrange_sample_by_taxa() %>%
    bar_plot(position = "fill")