Splitting the values in column using regex

Question

I have data.frame with two columns like the following

dat

    ID                             Details                         
    id_1        box1_homodomain gn=box1 os=homo sapiens p=4 se=1   
    id_2        sox2_plurinet gn=plu os=mus musculus p=5 se=3 

I would like to split the "os=xxx" and gn="yyy" in column "Details" for all the ids and print it like following:

    Id   Description        gn      os               
   Îd_1  box1_homodomain    box1    homo sapiens   
   Id_2  sox2_plurinet      plu     mouse musculus 

I tried the using gsub approach in R but I am unable to split the os=homo sapiens and gn=box1 into their respective columns. The following R code I used

dat$gn=gsub('^[gn=][A-z][A-z]`,dat$Details)
dat$os=gsub('^[os=][A-z][A-z]`,dat$Details)

Can anyone tell me what wrong and how can it be corrected. Kindly help me.

Thanks in advance

Answer 1

1) sub and gsub To do it using sub and gsub as in the question try this. Note that each regular expressions should match all of dat$Details so that when we replace it with the capture group only the capture group remains. For the dat$GO as in the comments to the question, we remove everything up to but not including P:, replace all occurrences of ;P with a comma and remove P: and also remove semicolon and everything thereafter. Similarly for F and C:

data.frame(dat[1], 
   Description = sub(" .*", "", dat$Details),
   gn = sub(".*gn=(.*) os=.*", "\\1", dat$Details),
   os = sub(".*os=(.*) p=.*", "\\1", dat$Details),
   P = gsub("P:|;.*", "", gsub(";P:", ",", sub(".*?P:", "P:", dat$GO))),
   F = gsub("F:|;.*", "", gsub(";F:", ",", sub(".*?F:", "F:", dat$GO))),
   C = gsub("C:|;.*", "", gsub(";C:", ",", sub(".*?C:", "C:", dat$GO))))

giving:

    ID     Description   gn           os       P       F       C
1 id_1 box1_homodomain box1 homo sapiens p_1,p_2     F_1 C_1,C_2
2 id_2   sox2_plurinet  plu mus musculus     p_1 F_1,F_2     C_1

2) read.pattern Processing of dat$Details is a bit easier with read.pattern (link) in the gsubfn package as one can define a single regular expression whose capture groups represent the fields of interest. Processing of dat$GO can be simplified too by extracting the P:... fields using strapplyc (link) and then concatenating them together with paste (and similarly with the F and C fields):

library(gsubfn)

Sub <- function(string, pat) sapply(strapplyc(string, pat), paste, collapse = ",")

DF <- read.pattern(text = as.character(dat$Details), 
        pattern = "(.*) gn=(.*) os=(.*) p=",
        col.names = c("Description", "gn", "os"),
        as.is = TRUE)

cbind(dat[1], DF,
      P = Sub(dat$GO, "P:(.*?);"),
      F = Sub(dat$GO, "F:(.*?);"),
      C = Sub(dat$GO, "C:(.*?);"))

giving:

    ID     Description   gn           os       P       F       C
1 id_1 box1_homodomain box1 homo sapiens p_1,p_2     F_1 C_1,C_2
2 id_2   sox2_plurinet  plu mus musculus     p_1 F_1,F_2     C_1

Here is a visualization of the regular expression used in read.pattern:

(.*) gn=(.*) os=(.*) p=

Debuggex Demo

Notes

1) If the dat$Details column is already character we could omit as.character. We could also omit as.is=TRUE if its ok to have factor columns in the result.

2) the sample output in the question has mouse but the input has mus. We have assumed it should be mus in both cases.

3) We used this for dat:

dat <-
structure(list(ID = c("id_1", "id_2"), 
Details = c("box1_homodomain gn=box1 os=homo sapiens p=4 se=1", 
"sox2_plurinet gn=plu os=mus musculus p=5 se=3"), 
GO = c("P:p_1;P:p_2;F:F_1;C:C_1;C:C_2;  ", 
"P:p_1;F:F_1;F:F_2;C:C_1;")), .Names = c("ID", "Details", 
"GO"), class = "data.frame", row.names = c(NA, -2L))

In the future please post the result of dput(dat) in the question.

Answer 2

You can also make use of regex capture groups for that. Each capture group match may be extracted with e.g. the stri_match_first_regex function from the stringi package.

dat <- data.frame(
   ID=c("id_1", "id_2"),
   details=c("box1_homodomain gn=box1 os=homo sapiens p=4 se=1", "sox2_plurinet gn=plu os=mus musculus p=5 se=3")
)

library(stringi)
res <- stri_match_first_regex(dat$details, "^(.+) gn=(.+) os=(.+) p=.*$")
res[,1] <- dat$ID
res <- as.data.frame(res)
names(res) <- c("ID", "Description", "gn", "os")
res
##   ID     Description   gn           os
## 1  1 box1_homodomain box1 homo sapiens
## 2  2   sox2_plurinet  plu mus musculus

Answer 3

Here's an option with tidyr:

library(tidyr)
# specify the new column names:
vars <- c("Description", "gn", "os")
# then separate the "Details" column according to regex and drop extra columns:
separate(dat, Details, into = vars, sep = "[A-Za-z]+=", extra = "drop")
#    ID      Description    gn            os
#1 id_1 box1_homodomain  box1  homo sapiens 
#2 id_2   sox2_plurinet   plu  mus musculus