How to manually set sample order in comp_barplot (when using the x argument)

[timz0605]

Hello David,

First of all, thank you so much for this amazing package!

I have a question about manually setting the order of bars in com_barplot. After some faceting and coding, even though I have created a list with the desired order for the bars to appear (given some geographical locations), R is still not understanding my command and is instead sorting the bars in descending order. I was wondering if you could provide any suggestions on this?

In the desired_order, I have a long list of row names I pulled from ps_sub, which has been arranged; they look like "M15E_24_4_1_16S_S43" "M15E_24_4_2_16S_S44" "M15E_24_4_3_16S_S45" "M1E_24_4_1_16S_S1" "M1E_24_4_2_16S_S2" "M1E_24_4_3_16S_S3".
For the Site_Rep from metadata, they look something like M15-1, M15-2, M15-3, M1-1, M1-2, M1-3. Instead of descending, ascending, or some random orders, I want them to be in the sequences of M15, M1, M2, M3, M4, M14, M4, M5 ... M13.

Below is the code I am using. I have also attached an image of the figure I currently have.

e_bar <- ps_sub %>% 
  tax_fix() %>% 
  comp_barplot(
    tax_level = "Family",
    x = "Site_Rep",
    label = "Site_Rep",
    sample_order = desired_order,
    n_taxa = 20,
    bar_width = 0.8,
    palette = met_tax_palette,
    bar_outline_color = NA,
    merge_other = TRUE
  ) +
  facet_grid(
    cols = vars(Season), 
    scales = "fixed"
    ) +
  coord_flip() +
  labs(x = "Site", 
       y = "Relative Abundance",
       title = "Dry Season Abundance, Ethanol Method") +
  guides(
    fill = guide_legend(ncol = 1)
  )

Thank you again!
Tim

[timz0605]

I think I have identified the potential issue? There seems to be something wrong with sample_order. Even though I tried to pre-arrange my phyloseq object and checked that the order is correct, and used sample_order = "asis" as the command, it is still not working and sorting my samples based on a descending order.

Some sorting I did:

# 1. Subset by method
ps_sub <- phyloseq %>% 
  subset_samples(Col_Method == "Ethanol") %>% 
  ps_mutate(
    Season = factor(Season, levels = c("Wet", "Dry")),
    Site   = factor(Site, levels   = station_levels)
  ) %>%
  ps_arrange(Site, .target = "sample_data") %>% 
  ps_arrange(Season, .target = "sample_data")

[david-barnett]

Hi ! 😄

The sample-based ordering is being ignored by comp_barplot because you are setting x = "Site_Rep", and so the categorical axis is sorted by the natural ordering of "Site_Rep" (alphabetical if it is a character variable)

If you first set Site_Rep to be a factor in the order you want them to appear, e.g. with ps_mutate, that should work without setting the sample_order argument in comp_barplot.

Does that make sense? the closest example to your situation in the docs is this one, but I don't currently explain this problem directly - so I'll update that in the docs at some point
https://david-barnett.github.io/microViz/articles/web-only/compositions.html#faceting-by-timepoint-and-another-variable

best
David