How to reduce aligned sequences to start and end coordinates?

Question

I have a data table of labelled coordinates that are aligned between two groups (A and B). For example:

dt_long <- data.table(LABEL_A = c(rep("A", 20), rep("A", 15), rep ("A", 10), rep ("A", 15), rep ("A", 10)),
                  SEQ_A = c(11:30, 61:75, 76:85, 86:100, 110:119),
                  LABEL_B= c(rep("C", 20), rep("D", 15), rep("F", 10), rep("G",15), rep("D", 10)),
                  SEQ_B = c(1:20, 25:11, 16:25, 15:1, 1:5, 8:12))

How can I reduce this information into a short format, where the start and end coordinates for each aligned sequence are given. For example:

dt_short <- data.table(LABEL_A = c("A", "A", "A", "A", "A", "A"),
                 Start_A = c(11, 61, 76, 86, 110, 115),
                 End_A = c(30, 75, 85, 100, 114, 119),
                 LABEL_B= c("C", "D", "F", "G", "D", "D"),
                 Start_B = c(1, 25, 16, 15, 1, 8),
                 End_B = c(20, 11, 25, 1, 5, 12))

The length of each aligned sequence should be identical. For example:

identical(abs(dt_short$End_A - dt_short$Start_A), abs(dt_short$End_B - dt_short$Start_B))

No, the label column distinguishes between sequences in different groups. — Powege, Jul 30 '19 at 14:07

chinsoon12 · Accepted Answer · 2019-08-01T00:13:36.487

1

You can make use of rleid and incorporating Frank's comment to remove grouping column

dt_long[, .(
        LABEL_A=LABEL_A[1L], Start_A=SEQ_A[1L], End_A=SEQ_A[.N], 
        LABEL_B=LABEL_B[1L], Start_B=SEQ_B[1L], End_B=SEQ_B[.N]), 
    by=rleid(LABEL_A, LABEL_B,
        c(0L, cumsum(diff(SEQ_A) > 1L)), 
        c(0L, cumsum(diff(SEQ_B) > 1L)))][, (1) := NULL]

output:

   LABEL_A Start_A End_A LABEL_B Start_B End_B
1:       A      11    30       C       1    20
2:       A      61    75       D      25    11
3:       A      76    85       F      16    25
4:       A      86   100       G      15     1
5:       A     110   114       D       1     5
6:       A     115   119       D       8    12

edited Aug 01 '19 at 00:13

answered Jul 31 '19 at 00:35

chinsoon12

25,005
4
25
35

Can this be adapted so that if either SEQ are not sequential they are defined by separate start and end coordinates? I have updated the example dataset to demonstrate this. Thanks. – Powege Jul 31 '19 at 07:46
addressed that as well – chinsoon12 Jul 31 '19 at 08:03
is this formatted correctly? It seems to be missing commas and brackets and is giving an error. – Powege Jul 31 '19 at 08:58
oops some copy paste problem – chinsoon12 Jul 31 '19 at 09:01
1

More efficient to drop a column with DT[, (1) := NULL] or something I guess. DT[, -1] creates a new table, I think. – Frank Jul 31 '19 at 17:54

score 0 · Answer 2 · answered Jul 30 '19 at 14:06

0

A straight forward way is to group by the two labels and get the first and last of each group, i.e.

library(data.table)

dt_long[, .(Start_A = first(SEQ_A), End_A = last(SEQ_A), Start_B = first(SEQ_B), End_B = last(SEQ_B)), by = .(LABEL_A, LABEL_B)][]

#   LABEL_A LABEL_B Start_A End_A Start_B End_B
#1:       1       3      11    30       1    20
#2:       1       4      61    75      25    11
#3:       1       6      76    85      16    25
#4:       1       7      86   100      15     1

answered Jul 30 '19 at 14:06

Sotos

51,121
6
32
66

Unfortunately, this does not work if there are multiple sequences with the same LABEL. I have now updated the examples. – Powege Jul 30 '19 at 14:41
Do you think an ID column could be added to group the aligned sequences? – Powege Jul 30 '19 at 19:21

score 0 · Answer 3 · answered Jul 30 '19 at 14:08

0

We can just subset and dcast. Would also work seamlessly when there are many different groups of columns

dcast(dt_long[, .SD[c(1, .N)], .(LABEL_A, LABEL_B)], 
  LABEL_A + LABEL_B ~ c("Start", "End")[rowid(LABEL_A, LABEL_B)], 
     value.var = c("SEQ_A", "SEQ_B"))
#    LABEL_A LABEL_B SEQ_A_End SEQ_A_Start SEQ_B_End SEQ_B_Start
#1:       1       3        30          11        20           1
#2:       1       4        75          61        11          25
#3:       1       6        85          76        25          16
#4:       1       7       100          86         1          15

answered Jul 30 '19 at 14:08

akrun

874,273
37
540
662

I get the following error: Error in .subset2(x, i, exact = exact) : subscript out of bounds – Powege Jul 30 '19 at 14:52
@Powege. Is it the error from the ssame example in your post, which I am not able to reproduce. I used `data.table_1.12.2` version – akrun Jul 30 '19 at 16:00
That is the error I get using the current example dataset and data.table version 1.10.4-3 – Powege Jul 30 '19 at 16:10
@Powege. Is it posssible for you to update the data.tablee version – akrun Jul 30 '19 at 16:30
Thanks, this now runs; however, it does produce the same dt_short as in the updated example. Can it be adapted to ensure the length of each aligned sequence are identical? – Powege Jul 30 '19 at 19:03

How to reduce aligned sequences to start and end coordinates?

3 Answers3