How can I use mutate() and case_when() in a for loop?

问题

I'm writing a Shiny app where the user will be inputting data for conditions of their samples, and the script will "automatically" match their inputted conditions to sample names of a given file.

For simplicity, I will not include the shiny code, because I am only struggling with the actual R implementation.

If I already know what the potential conditions are, I could do something like:

library(tidyverse)
x <- data.frame(Samples = c('Low1', 'Low2', 'High1', 'High2', 
                           'Ctrl1', 'Ctrl2'))

x <- x %>% mutate(Conditions = case_when(
           str_detect(Samples, fixed("low", ignore_case = T)) ~ "low",
           str_detect(Samples, fixed("high", ignore_case = T)) ~ "high",
           str_detect(Samples, fixed("ctrl", ignore_case = T)) ~ "ctrl"))

And I would get what I am looking for, a data frame like:

Samples    Conditions
   Low1           low
   Low2           low
  High1          high
  High2          high
  Ctrl1          ctrl
  Ctrl2          ctrl

However, I want to loop through a vector of potential conditions and do something like:

library(tidyverse)
condition_options <- c('low', 'high', 'ctrl')

x <- data.frame(Samples = samplenames)
for (j in condition_options) {
   x <- x %>% mutate(Condition = case_when(
        str_detect(Samples, fixed(j, ignore_case = T)) ~ j)) 
    }

When I do this, the Condition column is re-written only giving me matches for the last value in the vector. For example:

Samples    Conditions
   Low1         <NA>
   Low2         <NA>
  High1         <NA>
  High2         <NA>
  Ctrl1         ctrl
  Ctrl2         ctrl

回答1:

This might be easier if you build all parts of your case_when statement with meta-programming rather than doing a loop. Try

library(tidyverse)
condition_options <- c('low', 'high', 'ctrl')

conditions <- purrr::map(condition_options, 
                         ~quo(str_detect(Samples, fixed(!!.x, ignore_case = T))~!!.x))

x <- data.frame(Samples = samplenames)
x %>% mutate(Condition = case_when(!!!conditions) )

#   Samples Condition
# 1    Low1       low
# 2    Low2       low
# 3   High1      high
# 4   High2      high
# 5   Ctrl1      ctrl
# 6   Ctrl2      ctrl

Here the map build all the different formulas you would expect to have in the case_when statement. Then we use !!! to insert them into the mutate expression.

回答2:

library(purrr)
x <- data.frame(Samples = c('Low1', 'Low2', 'High1', 'High2', 
                            'Ctrl1', 'Ctrl2'))
condition_options <- c('low', 'high', 'ctrl')

# iterate through all provided `condition_options `, returns corresponding condition if a match is found, otherwise returns NA
matched_values <- map(condition_options,function(condition_name){
    ifelse(
        str_detect(x$Samples,fixed(condition_name,ignore_case = TRUE)),
        condition_name,
        NA_character_
    )
})

# if all values are NA, still return NA, otherwise return matched value, it will throw an error if multiple matches are found.
x["Conditions"] <- pmap_chr(values, function(...){
    values <- unlist(list(...))
    if(all(is.na(values))){
        return(NA)
    } else {
        return(values[!is.na(values)])
    }
})

> x
  Samples Conditions
1    Low1        low
2    Low2        low
3   High1       high
4   High2       high
5   Ctrl1       ctrl
6   Ctrl2       ctrl

回答3:

I don't think you'll need a loop to do this. We can use str_extract to extract any value which matches the pattern in condition_options

x$Conditions <- stringr::str_extract(tolower(x$Samples), 
                         paste0(condition_options, collapse = "|"))

x
#  Samples Conditions
#1    Low1        low
#2    Low2        low
#3   High1       high
#4   High2       high
#5   Ctrl1       ctrl
#6   Ctrl2       ctrl

In base R, we can also generate the regex dynamically using paste0

x$Conditions <- sub(paste0(".*(", paste0(condition_options, collapse = "|"), ").*"),
                "\\1", tolower(x$Samples))

where

paste0(".*(", paste0(condition_options, collapse = "|"), ").*") #gives
#[1] ".*(low|high|ctrl).*"

来源：https://stackoverflow.com/questions/57861055/how-can-i-use-mutate-and-case-when-in-a-for-loop

标签

tidyverse

mutate