一种选择是使用tidyr::separate_longer_delim（需要tidyr 1.3.0）和pivot_wider，如下所示：

library(tidyr)
library(dplyr)

dat |>
  mutate(row = row_number()) |> 
  separate_longer_delim(Classification, delim = ";") |> 
  mutate(suffix = toupper(substr(Classification, 1, 1))) |> 
  pivot_wider(names_from = suffix, values_from = Classification) |> 
  select(-row)
#> # A tibble: 5 × 6
#>   K           G               P                C               O          F    
#>   <chr>       <chr>           <chr>            <chr>           <chr>      <chr>
#> 1 k__Bacteria <NA>            <NA>             <NA>            <NA>       <NA> 
#> 2 k__Bacteria g__Streptofusia <NA>             <NA>            <NA>       <NA> 
#> 3 k__Bacteria <NA>            p__Lentisphaerae <NA>            <NA>       <NA> 
#> 4 k__Bacteria <NA>            p__Lentisphaerae c__Oligosphaeri o__Oligosp f__Ol
#> 5 k__Bacteria g__Ol           p__Lentisphaerae c__Oligosphaeri o__Oligosp f__Ol

数据

dat <- data.frame(
  Classification = c(
    "k__Bacteria",
    "k__Bacteria;g__Streptofusia", "k__Bacteria;p__Lentisphaerae",
    "k__Bacteria;p__Lentisphaerae;c__Oligosphaeri;o__Oligosp;f__Ol",
    "k__Bacteria;p__Lentisphaerae;c__Oligosphaeri;o__Oligosp;f__Ol;g__Ol"
  )
)

尝试separate_wider_regex，它非常适合这个用例。魔术发生在参数“patterns”上，它是一个可选命名元素的向量。未命名的元素被丢弃（在本例中是分号），但命名的元素被提取到带有名称的列中。

library(tidyr) #version 1.3.0

separate_wider_regex( data = dat,
                      col = Classification,
                      patterns = c(K = "k_+\\w+",
                                   ";",
                                   P = "p_+\\w+",
                                   ";",
                                   C = "c_+\\w+",
                                   ";",
                                   O = "o_+\\w+",
                                   ";",
                                   F = "f_+\\w+",
                                   ";",
                                   G = "g_+\\w+"
                                   ),
                      too_few = "align_start"
                      )

# A tibble: 5 × 6
  K           P                C               O       F     G    
  <chr>       <chr>            <chr>           <chr>   <chr> <chr>
1 k__Bacteria NA               NA              NA      NA    NA   
2 k__Bacteria NA               NA              NA      NA    NA   
3 k__Bacteria p__Lentisphaerae NA              NA      NA    NA   
4 k__Bacteria p__Lentisphaerae c__Oligosphaeri o__Oli… f__Ol NA   
5 k__Bacteria p__Lentisphaerae c__Oligosphaeri o__Oli… f__Ol g__Ol

• OBS：正如@stefan所指出的，这并不完全正确，因为当模式的顺序改变时，输出会丢失一些值。绝对值得展示separate_wider_regex thoguh，所以我在尝试修复它时会保留这个答案。