amr_upset() - MIC distribution panel - box plot middle line is not plotting the median?

[karatsang]

amr_upset() function

In regards to the MIC distribution panel that is generated by the amr_upset() function - the middle line within the box plot is not plotting the median MIC value?

Referring to the attached plot below. The code to reproduce panel a):

kp_mero <- download_ebi(
  antibiotic = "meropenem",
  species = "Klebsiella pneumoniae",
  reformat = TRUE,
  interpret_eucast = TRUE,
  interpret_ecoff = TRUE
)
kp_mero_euscape <- kp_mero %>% filter(grepl("31358985", source))

contaminated_assemblies <- c("SAMEA3729690", "SAMEA3721062", "SAMEA3721052", "SAMEA3720966", "SAMEA3673128", "SAMEA3538742", "SAMEA3721188", "SAMEA3649589", "SAMEA3538652", "SAMEA3649503", "SAMEA3538911", "SAMEA3727711", "SAMEA3649452", "SAMEA3649453", "SAMEA3649454", "SAMEA3649467", "SAMEA3721063", "SAMEA3538862", "SAMEA3538667", "SAMEA3673004", "SAMEA3729818", "SAMEA3729660", "SAMEA3673078", "SAMEA3673097")

kp_mero_euscape <- kp_mero_euscape %>%
  filter(!id %in% contaminated_assemblies)

kleborate_dev <- import_kleborate(kleborate_raw)


kp_mero_kleborate_binary_matrix <- get_binary_matrix(
  geno_table = kleborate_dev,
  pheno_table = kp_mero_euscape,
  antibiotic = "Meropenem",
  drug_class_list = c("Carbapenems"),
  sir_col = "pheno_eucast",
  keep_assay_values = TRUE,
  keep_assay_values_from = "mic",
  marker_col = "marker.label"
)

kp_mic_upset_kleborate <- amr_upset(kp_mero_kleborate_binary_matrix, assay = "mic", bp_R="8", bp_S="2", ecoff_bp="0.125", min_set_size = 1)

kp_mic_upset_kleborate$plot
kp_mic_upset_kleborate$summary

Reviewing kp_mic_upset_kleborate$summary, I am not sure if the box plot middle line is plotting median_excludeRangeValues or median_ignoreRanges? For example, in panel a) in the no markers (first column) of the UpSet plot, the median looks to be 1.5mg/L, but the median_excludeRangeValues= 0.25 and median_ignoreRanges = 0.06 from kp_mic_upset_kleborate$summary.

As a "sanity check", in panel b) - in the no carbapenem, wt porin (first) column, it's plotting the same data using geom_boxplot(width = 0.7, outlier.shape = NA) and the median is <0.25mg/L

I haven't looked too deeply into this, but will investigate tomorrow

[katholt]

Hi @karatsang, the boxplots are plotted using geom_boxplot() with aes(y=mic) (or disk), so the middle line will be the median of the values in that column as calculated by geom_boxplot(), not the medians with/without ranges that we calculate directly and report in the summary table. I would expect that to equate to the 'ignoreRanges' median we calculate, but it may be doing something unexpected.
You raise a good point and we should try to make it so the boxplots show the median calculated by the function, ideally the choice of which flavour, not some third value as that is indeed confusing!