我们可以在行上使用base R-loop，用na.omit移除NA，用n = 2和paste获得最后两个元素tail

demo_table$assignation <- apply(demo_table, 1, 
   function(x) paste(tail(na.omit(x), 2), collapse = " "))

• 输出

demo_table$assignation
[1] "UBA4441 UBA4441 sp"            "Cyanobacteria Cyanobacteriia"  "Phycisphaerae Phycisphaerales" "Poseidoniales Poseidonia"     
[5] "Archaea Thermoplasmatota"

对于tidyverse，我们也可以使用unite，并使用na.rm = TRUE移除NA，然后提取最后两个字

library(dplyr)
library(tidyr)
library(stringr)
demo_table %>% 
  unite(assignation, everything(), na.rm = TRUE, remove = FALSE) %>% 
  mutate(assignation = str_replace(assignation,     
     ".*_([^_]+)_([^_]+)$", "\\1 \\2")) %>% 
  relocate(assignation, .after =last_col())

如果你想使用tidyverse方法，你只需要使用rowwise + c_across。我认为把这个操作转换成一个函数也很好，以防你以后需要使用或者甚至改变它的行为。
编号

library(dplyr)

select_last <- function(x, n = 2){paste(tail(na.omit(x),n = n),collapse = " ")}

demo_table %>% 
  rowwise() %>% 
  mutate(assignation  = select_last(c_across()))

输出

# A tibble: 5 x 8
# Rowwise: 
  Domain  Phylum      Class        Order       Family     Genus Species  assignation        
  <chr>   <chr>       <chr>        <chr>       <chr>      <chr> <chr>    <chr>              
1 Bacter~ Proteobact~ Alphaproteo~ Sphingomon~ Emcibacte~ UBA4~ UBA4441~ UBA4441 UBA4441 sp 
2 Bacter~ Cyanobacte~ Cyanobacter~ NA          NA         NA    NA       Cyanobacteria Cyan~
3 Bacter~ Planctomyc~ Phycisphaer~ Phycisphae~ NA         NA    NA       Phycisphaerae Phyc~
4 Archaea Thermoplas~ Poseidoniia~ Poseidonia~ Poseidonia NA    NA       Poseidoniales Pose~
5 Archaea Thermoplas~ NA           NA          NA         NA    NA       Archaea Thermoplas~

以下是dplyr与tidyr结合使用的方法：

library(dplyr)
library(tidyr)
demo_table %>% 
  mutate(id = row_number()) %>% 
  pivot_longer(-id) %>% 
  group_by(id) %>% 
  na.omit() %>% 
  arrange(-row_number(), .by_group = TRUE) %>% 
  mutate(assignation = paste(value[1], value[2], sep = "\n")) %>% 
  slice(1) %>% 
  ungroup() %>% 
  select(assignation) %>% 
  bind_cols(demo_table) %>% 
  View()