replacing rowwise() operations in grouped data - r

Question

Anonymised example subset of a much larger dataset (now edited to show an option with multiple competing types):
structure(list(`Sample File` = c("A", "A", "A", "A", "A", "A",
"A", "A", "A", "B", "B", "B", "B", "B", "C", "C", "C", "C"),
Marker = c("X", "X", "X", "X", "Y", "Y", "Y", "Y", "Y", "Z",
"Z", "Z", "Z", "Z", "q", "q", "q", "q"), Allele = c(19, 20,
22, 23, 18, 18.2, 19, 19.2, 20, 12, 13, 14, 15, 16, 10, 10.2,
11, 12), Size = c(249.15, 253.13, 260.64, 264.68, 366, 367.81,
369.97, 372.02, 373.95, 91.65, 95.86, 100, 104.24, 108.38,
177.51, 179.4, 181.42, 185.49), Height = c(173L, 1976L, 145L,
1078L, 137L, 62L, 1381L, 45L, 1005L, 38L, 482L, 5766L, 4893L,
19L, 287L, 36L, 5001L, 50L), Type = c("minusone", "allele",
"minusone", "allele", "ambiguous", "minushalf", "allele",
"minushalf", "allele", "minustwo", "ambiguous", "allele",
"allele", "plusone", "minusone", "minushalf", "allele", "plusone"
), LUS = c(11.75, 11.286, 13.375, 13.5, 18, 9, 19, 10, 20,
12, 11, 14, 15, 16, 9.5, NA, 11, 11.5)), class = c("grouped_df",
"tbl_df", "tbl", "data.frame"), row.names = c(NA, -18L), groups = structure(list(
`Sample File` = c("A", "A", "B", "C"), Marker = c("X", "Y",
"Z", "q"), .rows = structure(list(1:4, 5:9, 10:14, 15:18), ptype = integer(0), class = c("vctrs_list_of",
"vctrs_vctr", "list"))), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -4L), .drop = TRUE))
I want to look up values based on the classification $Type.
"minustwo" means I want to look up the "Allele", "Height" and "LUS"
values for the row with "Allele" equal to the current row plus two,
with the same Sample File and Marker.
"minusone" means the same but for "Allele" equal to the current row plus one.
"minushalf" means the same but for "Allele" equal to the current row plus 0.2 but the dot values here are 25% each, so 12.1, 12.3, 12.3, 13, 13.1 etc - I have a helper function plusTwoBP() for this.
"plusone" means the same for "Allele" equal to the current row -1
"allele" or "ambiguous" don't need to do anything.
Ideal output:
# A tibble: 18 × 10
# Rowwise: Sample File, Marker
`Sample File` Marker Allele Size Height Type LUS ParentHeight ParentAllele ParentLUS
<chr> <chr> <dbl> <dbl> <int> <chr> <dbl> <int> <dbl> <dbl>
1 A X 19 249. 173 minusone 11.8 1976 20 11.3
2 A X 20 253. 1976 allele 11.3 NA NA NA
3 A X 22 261. 145 minusone 13.4 1078 23 13.5
4 A X 23 265. 1078 allele 13.5 NA NA NA
5 A Y 18 366 137 ambiguous 18 NA NA NA
6 A Y 18.2 368. 62 minushalf 9 1381 19 19
7 A Y 19 370. 1381 allele 19 NA NA NA
8 A Y 19.2 372. 45 minushalf 10 1005 20 20
9 A Y 20 374. 1005 allele 20 NA NA NA
10 B Z 12 91.6 38 minustwo 12 5766 14 14
11 B Z 13 95.9 482 ambiguous 11 NA NA NA
12 B Z 14 100 5766 allele 14 NA NA NA
13 B Z 15 104. 4893 allele 15 NA NA NA
14 B Z 16 108. 19 plusone 16 4893 15 15
15 C q 10 178. 287 minusone 9.5 5001 11 11
16 C q 10.2 179. 36 minushalf NA 5001 11 11
17 C q 11 181. 5001 allele 11 NA NA NA
18 C q 12 185. 50 plusone 11.5 5001 11 11
I have a rather belaboured way of doing it:
# eg for minustwo
sampleData %>%
filter(Type == "minustwo") %>%
rowwise() %>%
mutate(ParentHeight = sampleData$Height[sampleData$`Sample File` == `Sample File` & sampleData$Marker == Marker & sampleData$Allele == (Allele + 2)],
ParentAllele = sampleData$Allele[sampleData$`Sample File` == `Sample File` & sampleData$Marker == Marker & sampleData$Allele == (Allele + 2)],
ParentLUS = sampleData$LUS[sampleData$`Sample File` == `Sample File` & sampleData$Marker == Marker & sampleData$Allele == (Allele + 2)]) %>%
right_join(sampleData)
I then have to redo that for each of my Types
My real dataset is thousands of rows so this ends up being a little slow but manageable, but more to the point I want to learn a better way to do it, in particular the sampleData$'Sample File' == 'Sample File' & sampleData$Marker == Marker seems like it should be doable with grouping so I must be missing a trick there.
I have tried using group_map() but I've clearly not understood it correctly:
sampleData$ParentHeight <- sampleData %>%
group_by(`Sample File`, `Marker`) %>%
group_map(.f = \(.x, .y) {
pmap_dbl(.l = .x, .f = \(Allele, Height, Type, ...){
if(Type == "allele" | Type == "ambiguous") { return(0)
} else if (Type == "plusone") {
return(.x$Height[.x$Allele == round(Allele - 1, 1)])
} else if (Type == "minushalf") {
return(.x$Height[.x$Allele == round(plustwoBP(Allele), 1)])
} else if (Type == "minusone") {
return(.x$Height[.x$Allele == round(Allele + 1, 1)])
} else if (Type == "minustwo") {
return(.x$Height[.x$Allele == round(Allele + 2, 1)])
} else { stop("unexpected peak type") }
})}) %>% unlist()
Initially seems to work, but on investigation it's not respecting both layers of grouping, so brings matches from the wrong Marker. Additionally, here I'm assigning the output to a new column in the data frame, but if I try to instead wrap a mutate() around this so that I can create all three new columns in one go then the group_map() no longer works at all.
I also considered using complete() to hugely extend the data frame will all possible values of Allele (including x.0, x.1, x.2, x.3 variants) then use lag() to select the corresponding rows, then drop the spare rows. This seems like it'd make the data frame enormous in the interim.
To summarise
This works, but it feels ugly and like I'm missing a more elegant and obvious solution. How would you approach this?

Answer 1

You can create two versions of Allele: one identical to the original Allele, and one that is equal to an adjustment based on minusone, minustwo, etc
Then do a self left join, based on that adjusted version of Allele (and Sample File and Marker)
sampleData = sampleData %>% group_by(`Sample File`,Marker) %>% mutate(id = Allele) %>% ungroup()
left_join(
sampleData %>%
mutate(id = case_when(
Type=="minusone"~id+1,
Type=="minustwo"~id+2,
Type=="plusone"~id-1,
Type=="minushalf"~ceiling(id))),
sampleData %>% select(-c(Size,Type)),
by=c("Sample File", "Marker", "id"),
suffix = c("", ".parent")
) %>% select(-id)
Output:
# A tibble: 14 × 10
`Sample File` Marker Allele Size Height Type LUS Allele.parent Height.parent LUS.parent
<chr> <chr> <dbl> <dbl> <int> <chr> <dbl> <dbl> <int> <dbl>
1 A X 19 249. 173 minusone 11.8 20 1976 11.3
2 A X 20 253. 1976 allele 11.3 NA NA NA
3 A X 22 261. 145 minusone 13.4 23 1078 13.5
4 A X 23 265. 1078 allele 13.5 NA NA NA
5 A Y 18 366 137 ambiguous 18 NA NA NA
6 A Y 18.2 368. 62 minushalf 9 19 1381 19
7 A Y 19 370. 1381 allele 19 NA NA NA
8 A Y 19.2 372. 45 minushalf 10 20 1005 20
9 A Y 20 374. 1005 allele 20 NA NA NA
10 B Z 12 91.6 38 minustwo 12 14 5766 14
11 B Z 13 95.9 482 ambiguous 11 NA NA NA
12 B Z 14 100 5766 allele 14 NA NA NA
13 B Z 15 104. 4893 allele 15 NA NA NA
14 B Z 16 108. 19 plusone 16 15 4893 15
15 C q 10 178. 287 minusone 9.5 11 5001 11
16 C q 10.2 179. 36 minushalf NA 11 5001 11
17 C q 11 181. 5001 allele 11 NA NA NA
18 C q 12 185. 50 plusone 11.5 11 5001 11