How can I specify pairwise comparisons of interest rather than computing all possible comparisons with something like Tukey?

Question

Here is a subset of my df. How can I run pairwise comparisons of median expression between disease status for each clus_ft and use Bonferroni correction?

For example, from this df subset, I would want to perform the following comparisons:

1. A_CD1, no disease vs. A_CD1, with disease
2. B_CD1, no disease vs. B_CD1, with disease
3. A_IFN, no disease vs. A_IFN, with disease

I do not want to compute pvalues for all possible pairwise comparisons. For example, the following comparison is irrelevant and should not be computed:

A_CD1, no disease vs. A_IFN, with disease

| patient_id | median | disease | clus_ft |
| ---------- | ------ | ------- | ------- |
| 1          | 0.40   | no      | A_CD1   |
| 2          | 0.29   | no      | A_CD1   | 
| 1          | 0.43   | yes     | A_CD1   | 
| 2          | 0.32   | yes     | A_CD1   |
| 1          | 0.77   | no      | B_CD1   | 
| 2          | 0.65   | no      | B_CD1   |
| 1          | 0.01   | yes     | B_CD1   | 
| 2          | 0.06   | yes     | B_CD1   | 
| 1          | 0.86   | no      | A_IFN   | 
| 2          | 0.85   | no      | A_IFN   |
| 1          | 0.02   | yes     | A_IFN   | 
| 2          | 0.24   | yes     | A_IFN   | 
Showing 1 to 12 of 960 entries

Answer 1

You can use dplyr to extract relevant vectors of values for this comparison. Build out from there using, say, a vector of clus_ft values.

Step-1

## untested
run_t_test <- function(df, in_clus_ft) {
  t.test(x = df %>% filter(clus_ft == in_clus_ft & disease == "yes") %>% pull(median),
         y = df %>% filter(clus_ft == in_clus_ft & disease == "no") %>% pull(median))
}

run_t_test(df, "A_CD1")

Step-2: Now run it on a set of values

sapply(X = c("A_CD1","B_CD1","A_IFN"),
FUN = run_t_test,
df = df
)