significant associations of output and boxplot?

[pandaqiuqiu]

Hi,

After I used：
res = seu |>
sccomp_estimate(formula_composition = ~ type,
.sample = sample,
.cell_group = cell_group,
bimodal_mean_variability_association = TRUE) |>
sccomp_test()

the result for res is as follows:

According to the results, c_FDR for the cancer factor is 0.603. However, the boxplot generated by res |>
sccomp_boxplot(factor = "type")
shows a significant difference with a red box.

I would like to ask:
Isn't the definition of significant difference indicated by c_FDR < 0.05? Why is my result c_FDR = 0.603, yet the boxplot is red?

Thanks!

[stemangiola]

I think something to do with visualisation. Would you mind send me the count dataset (anonimised if needed) so I will be able to replicate your scenario and fix it?

Thanks.

[pandaqiuqiu]

I think something to do with visualisation. Would you mind send me the count dataset (anonimised if needed) so I will be able to replicate your scenario and fix it?

Thanks.

The count data is here. Thanks for your rapid response.
sccomp--.csv

[stemangiola]

Hello,

you might have an old version of sccomp (btw, now the default fold change used for testing is 0.1, before was 0.2, as we deemed more appropriate).

Please reinstall sccomp from github

You still not have significance. The plot reflects this

    read_csv("~/Downloads/sccomp--.csv") |>
    mutate(count = count |> as.integer()) |> 
    sccomp_estimate(formula_composition = ~ Group0,
                    .sample = Sample,
                    .cell_group = cell_group,.abundance = count, inference_method = "hmc",
                    bimodal_mean_variability_association = TRUE) |>
    sccomp_remove_outliers(inference_method = "hmc") |> 
    sccomp_test() |> 
        sccomp_boxplot("Group0")