Dplyr: group_by and convert multiple columns to a vector - r

Question

I have a question about how to convert multiple columns to a vector. I have the following dataset that I would like to group them by their condition and take all the position count into one vector. I know I can use as.vector() to convert them individually but I wonder if there is a dplyr way. Thank you!
test -> structure(list(gene_id = c("gene0", "gene0", "gene0", "gene0",
"gene0", "gene0", "gene0", "gene0", "gene0", "gene0", "gene0",
"gene0", "gene0", "gene0", "gene0", "gene0", "gene0", "gene0",
"gene0", "gene0", "gene0", "gene0", "gene0", "gene0"), codon_index = c(1L,
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L,
3L, 1L, 2L, 3L, 1L, 2L, 3L), position_1_count = c(2L, 7L, 8L,
0L, 2L, 22L, 19L, 15L, 134L, 1L, 127L, 30L, 0L, 0L, 1L, 4L, 65L,
234L, 1L, 3L, 57L, 0L, 4L, 16L), position_2_count = c(0L, 5L,
5L, 0L, 3L, 2L, 3L, 13L, 134L, 0L, 36L, 5L, 0L, 0L, 0L, 1L, 150L,
7L, 0L, 7L, 7L, 0L, 6L, 1L), position_3_count = c(0L, 2L, 1L,
0L, 4L, 0L, 3L, 32L, 43L, 3L, 9L, 1L, 0L, 0L, 0L, 4L, 105L, 1L,
0L, 14L, 5L, 0L, 6L, 1L), condition = structure(c(1L, 1L, 1L,
7L, 7L, 7L, 3L, 3L, 3L, 5L, 5L, 5L, 8L, 8L, 8L, 2L, 2L, 2L, 4L,
4L, 4L, 6L, 6L, 6L), .Label = c("c", "cup", "n", "nup", "p",
"pup", "min", "rich"), class = "factor")), class = c("tbl_df",
"tbl", "data.frame"), row.names = c(NA, -24L), .Names = c("gene_id",
"codon_index", "position_1_count", "position_2_count", "position_3_count",
"condition"))
> head(a)
# A tibble: 6 × 6
gene_id codon_index position_1_count position_2_count position_3_count condition
<chr> <int> <int> <int> <int> <fctr>
1 gene0 1 2 0 0 c
2 gene0 2 7 5 2 c
3 gene0 3 8 5 1 c
4 gene0 1 0 0 0 min
5 gene0 2 2 3 4 min
6 gene0 3 22 2 0 min
How can we convert this dataset to (I didn't add the column names here)
2 0 0 7 5 2 8 5 1 c
0 0 0 2 3 4 22 2 0 min

Answer 1

Another alternative:
library(purrr)
test %>%
slice_rows("condition") %>%
by_slice(function(x) unlist(x[-(1:2)]), .to = "vec")
Which gives:
# condition vec
#1 c 2, 7, 8, 0, 5, 5, 0, 2, 1
#2 cup 4, 65, 234, 1, 150, 7, 4, 105, 1
#3 n 19, 15, 134, 3, 13, 134, 3, 32, 43
#4 nup 1, 3, 57, 0, 7, 7, 0, 14, 5
#5 p 1, 127, 30, 0, 36, 5, 3, 9, 1
#6 pup 0, 4, 16, 0, 6, 1, 0, 6, 1
#7 min 0, 2, 22, 0, 3, 2, 0, 4, 0
#8 rich 0, 0, 1, 0, 0, 0, 0, 0, 0
As per mentioned in the comments by #advance, if you want the result rowwise:
test %>%
slice_rows("condition") %>%
by_slice(function(x) as.vector(t(x[-(1:2)])), .to = "vec")
# condition vec
#1 c 2, 0, 0, 7, 5, 2, 8, 5, 1
#2 cup 4, 1, 4, 65, 150, 105, 234, 7, 1
#3 n 19, 3, 3, 15, 13, 32, 134, 134, 43
#4 nup 1, 0, 0, 3, 7, 14, 57, 7, 5
#5 p 1, 0, 3, 127, 36, 9, 30, 5, 1
#6 pup 0, 0, 0, 4, 6, 6, 16, 1, 1
#7 min 0, 0, 0, 2, 3, 4, 22, 2, 0
#8 rich 0, 0, 0, 0, 0, 0, 1, 0, 0
Or adapting #DavidArenburg's comment using do() instead of summarise():
test %>%
group_by(condition) %>%
select(position_1_count:condition) %>%
do(res = c(t(.[,-4])))
Which gives:
# condition res
#1 c 2, 0, 0, 7, 5, 2, 8, 5, 1
#2 cup 4, 1, 4, 65, 150, 105, 234, 7, 1
#3 n 19, 3, 3, 15, 13, 32, 134, 134, 43
#4 nup 1, 0, 0, 3, 7, 14, 57, 7, 5
#5 p 1, 0, 3, 127, 36, 9, 30, 5, 1
#6 pup 0, 0, 0, 4, 6, 6, 16, 1, 1
#7 min 0, 0, 0, 2, 3, 4, 22, 2, 0
#8 rich 0, 0, 0, 0, 0, 0, 1, 0, 0

Answer 2

Am I correct that what you want is a separate vector for all of the counts for each condition? If so, a mix of dplyr and tidyr should do it. First, I gather to put all of the counts in a single column. Then, split to separate by the condition, then use lapply to generate a list, containing a separate vector for each condition:
a %>%
gather(Location, Count, starts_with("position")) %>%
split(.$condition) %>%
lapply(function(x){x$Count})
gives:
$c
[1] 2 7 8 0 5 5 0 2 1
$cup
[1] 4 65 234 1 150 7 4 105 1
$n
[1] 19 15 134 3 13 134 3 32 43
$nup
[1] 1 3 57 0 7 7 0 14 5
$p
[1] 1 127 30 0 36 5 3 9 1
$pup
[1] 0 4 16 0 6 1 0 6 1
$min
[1] 0 2 22 0 3 2 0 4 0
$rich
[1] 0 0 1 0 0 0 0 0 0
If the order matters (and is wrong above) you should be able to sort before splitting, e.g. by adding arrange(codon_index) after gather

Answer 3

After taking Peterson's idea, I think this code works the best:
test %>% gather(Location, Count, starts_with("position")) %>% arrange(codon_index) %>% group_by(condition) %>% do(count = as.vector(t(.$Count)))
The result will look like this
> ans = test %>% gather(Location, Count, starts_with("position")) %>% arrange(codon_index) %>% group_by(condition) %>% do(count = as.vector(t(.$Count)))
# A tibble: 8 × 2
condition count
* <fctr> <list>
1 c <int [9]>
2 cup <int [9]>
3 n <int [9]>
4 nup <int [9]>
5 p <int [9]>
6 pup <int [9]>
7 min <int [9]>
8 rich <int [9]>
> ans$count[[1]]
[1] 2 0 0 7 5 2 8 5 1
Thanks a lot for everyone's help!