问题
I am new to R. I'm trying to filter rows from a data.frame (df) based on multiple conditions:
An example of my data.frame: image of my df
df:
SNPA SNPB value block1 block2 score_T
A1 A22 0.379927 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 12
A2 A23 0.449074 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 25
A3 A24 0.464135 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 584
A4 A22 0.328866 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 51
A5 A22 0.326026 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 64
A22 A27 0.57169 A22|A23|A24|A25 A27|A28|A29|A30|A31 77
A23 A28 0.416178 A22|A23|A24|A25 A27|A28|A29|A30|A31 90
A24 A29 0.456144 A22|A23|A24|A25 A27|A28|A29|A30|A31 103
A34 A39 0.379927 A31|A32|A33|A34 A39|A40|A41|A42 116
A34 A40 0.759074 A31|A32|A33|A34 A39|A40|A41|A42 129
A34 A41 0.562303 A31|A32|A33|A34 A39|A40|A41|A42 142
A39 A57 0.322303 A39|A40|A41|A42 A52|A53|A54|A55|A56|A57|A58|A59|A60|A61 25
A40 A57 0.372303 A39|A40|A41|A42 A52|A53|A54|A55|A56|A57|A58|A59|A60|A61 198
A41 A57 0.562303 A39|A40|A41|A42 A52|A53|A54|A55|A56|A57|A58|A59|A60|A61 356
What I want is to use dplyr to keep only the rows where the blocks (block1 and block2) have at least two SNPs (from the column SNPA for block1, and SNPB for block2) each, and remove pairs of blocks containing 1 SNP (example: rows 9 to 14).
Result wanted: result
SNPA SNPB value block1 block2 score_T
A1 A22 0.379927 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 12
A2 A23 0.449074 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 25
A3 A24 0.464135 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 584
A4 A22 0.328866 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 51
A5 A22 0.326026 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 64
A22 A27 0.57169 A22|A23|A24|A25 A27|A28|A29|A30|A31 77
A23 A28 0.416178 A22|A23|A24|A25 A27|A28|A29|A30|A31 90
A24 A29 0.456144 A22|A23|A24|A25 A27|A28|A29|A30|A31 103
Do you have any idea how I can do that?
result <- df %>% group_by(block1, block2) %>% filter(...) %>% summarise(mean_s = mean(score_T), number = n())
Thanks.
回答1:
A somewhat slow base-dplyr solution. Some issues with this solution include the need to manually set "blocks" and "snps" in our filter function. One could probably automate this process.
my_filter <- function(df,block, snp){
res<-strsplit(df[[block]],"|", fixed= TRUE)
lengths(lapply(res, function(x) which(x %in% df[[snp]]))) > 1
}
df %>%
filter(my_filter(., "block1", "SNPA"), my_filter(., "block2","SNPB"))
SNPA SNPB value block1 block2 score_T
1 A1 A22 0.379927 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 12
2 A2 A23 0.449074 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 25
3 A3 A24 0.464135 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 584
4 A4 A22 0.328866 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 51
5 A5 A22 0.326026 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 64
6 A22 A27 0.571690 A22|A23|A24|A25 A27|A28|A29|A30|A31 77
7 A23 A28 0.416178 A22|A23|A24|A25 A27|A28|A29|A30|A31 90
8 A24 A29 0.456144 A22|A23|A24|A25 A27|A28|A29|A30|A31 103
Data:
df <-structure(list(SNPA = c("A1", "A2", "A3", "A4", "A5", "A22",
"A23", "A24", "A34", "A34", "A34", "A39", "A40", "A41"), SNPB = c("A22",
"A23", "A24", "A22", "A22", "A27", "A28", "A29", "A39", "A40",
"A41", "A57", "A57", "A57"), value = c(0.379927, 0.449074, 0.464135,
0.328866, 0.326026, 0.57169, 0.416178, 0.456144, 0.379927, 0.759074,
0.562303, 0.322303, 0.372303, 0.562303), block1 = c("A1|A2|A3|A4|A5|A6",
"A1|A2|A3|A4|A5|A6", "A1|A2|A3|A4|A5|A6", "A1|A2|A3|A4|A5|A6",
"A1|A2|A3|A4|A5|A6", "A22|A23|A24|A25", "A22|A23|A24|A25", "A22|A23|A24|A25",
"A31|A32|A33|A34", "A31|A32|A33|A34", "A31|A32|A33|A34", "A39|A40|A41|A42",
"A39|A40|A41|A42", "A39|A40|A41|A42"), block2 = c("A22|A23|A24|A25",
"A22|A23|A24|A25", "A22|A23|A24|A25", "A22|A23|A24|A25", "A22|A23|A24|A25",
"A27|A28|A29|A30|A31", "A27|A28|A29|A30|A31", "A27|A28|A29|A30|A31",
"A39|A40|A41|A42", "A39|A40|A41|A42", "A39|A40|A41|A42", "A52|A53|A54|A55|A56|A57|A58|A59|A60|A61",
"A52|A53|A54|A55|A56|A57|A58|A59|A60|A61", "A52|A53|A54|A55|A56|A57|A58|A59|A60|A61"
), score_T = c(12L, 25L, 584L, 51L, 64L, 77L, 90L, 103L, 116L,
129L, 142L, 25L, 198L, 356L)), class = "data.frame", row.names = c(NA,
-14L))
回答2:
toString(unique(SNPA)) will return a character vector of length one, which will be matched with block1 using setdiff, then we apply this process to each row using purrr::map2
#Here a toy example to understand setdiff, strsplit, and length
> length(strsplit(setdiff(toString(unique(df[1:5,'SNPA'])), df[1,'block1']),',')[[1]])
[1] 5
library(dplyr)
library(purrr)
df %>%
group_by(block1,block2) %>%
mutate(A = purrr::map2_dbl(toString(unique(SNPA)), block1, ~length(strsplit(setdiff(.x, .y),',')[[1]])),
B = purrr::map2_dbl(toString(unique(SNPB)), block2, ~length(strsplit(setdiff(.x, .y),',')[[1]]))) %>%
filter(A>2 & B>2)
# A tibble: 8 x 8
# Groups: block1, block2 [2]
SNPA SNPB value block1 block2 score_T A B
<chr> <chr> <dbl> <chr> <chr> <int> <dbl> <dbl>
1 A1 A22 0.380 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 12 5 3
2 A2 A23 0.449 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 25 5 3
3 A3 A24 0.464 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 584 5 3
4 A4 A22 0.329 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 51 5 3
5 A5 A22 0.326 A1|A2|A3|A4|A5|A6 A22|A23|A24|A25 64 5 3
6 A22 A27 0.572 A22|A23|A24|A25 A27|A28|A29|A30|A31 77 3 3
7 A23 A28 0.416 A22|A23|A24|A25 A27|A28|A29|A30|A31 90 3 3
8 A24 A29 0.456 A22|A23|A24|A25 A27|A28|A29|A30|A31 103 3 3
来源：https://stackoverflow.com/questions/58559555/how-to-filter-rows-with-multiple-conditions