根据逗号分隔的字符向量列的值熔合表（data.frame）

Question

I'm doing an experiment where I have "regions" with some associated statistic (actually many other statistics and descriptive columns), and a comma-separated list of genes that lie in those regions. This list will be variable in number, and may not contain anything ("NA").

How can I "melt" table a:

  region_id  statistic      genelist
          1        2.5       A, B, C
          2        0.5    B, C, D, E
          3        3.2          <NA>
          4        0.1          E, F

To create another table with a separate entry for each gene in the list of genes? Ie

   region_id statistic gene
           1       2.5    A
           1       2.5    B
           1       2.5    C
           2       0.5    B
           2       0.5    C
           2       0.5    D
           2       0.5    E
           3       3.2 <NA>
           4       0.1    E
           4       0.1    F

I'm guessing there's a way to do this with R/plyr, but I'm not sure how. Thanks in advance.

Edit:

Using R you can recreate these toy vectors with this code:

a <- structure(list(region_id = 1:4, statistic = c(2.5, 0.5, 3.2, 
0.1), genelist = structure(c(1L, 2L, NA, 3L), .Label = c("A, B, C", 
"B, C, D, E", "E, F"), class = "factor")), .Names = c("region_id", 
"statistic", "genelist"), class = "data.frame", row.names = c(NA, 
-4L))

b <- structure(list(region_id = c(1L, 1L, 1L, 2L, 2L, 2L, 2L, 3L, 
4L, 4L), statistic = c(2.5, 2.5, 2.5, 0.5, 0.5, 0.5, 0.5, 3.2, 
0.1, 0.1), gene = structure(c(1L, 2L, 3L, 2L, 3L, 4L, 5L, NA, 
5L, 6L), .Label = c("A", "B", "C", "D", "E", "F"), class = "factor")), .Names = c("region_id", 
"statistic", "gene"), class = "data.frame", row.names = c(NA, 
-10L))

Answer 1

A data.table solution for time, memory and coding efficiency

library(data.table)
DT <- data.table(a)
DT[, list(statistic, 
          gene = unlist(strsplit(as.character(genelist), ', ' ))),
   by = list(region_id)]

Or you could use the nice formatting of of list from data.table version >= 1.8.2

DTL <- DT[, list(statistic, 
         gene = strsplit(as.character(genelist), ', ' )),
    by = list(region_id)]

DTL
##    region_id statistic    gene
## 1:         1       2.5   A,B,C
## 2:         2       0.5 B,C,D,E
## 3:         3       3.2      NA
## 4:         4       0.1     E,F

In which case gene is a list of lists

DTL[region_id == 1,unlist(gene)]
## [1] "A" "B" "C"
DTL[region_id == 2,unlist(gene)]
## [1] "B" "C" "D" "E"
# or if the following is of interest
DTL[statistic < 2,unlist(gene)]
## [1] "B" "C" "D" "E" "E" "F"

etc

Answer 2

Simply split the fields, then split the genes and print one line per gene. You can try this out in a script by replacing <DATA> with <> and using the input file as argument to the perl script, eg perl script.pl input.txt .

use strict;
use warnings;

while (<DATA>) {
    chomp;                                   # remove newline
    my ($reg, $stat, $gene) = split /\t/;    # split fields
    my @genes = split /,\s*/, $gene;         # split genes
    for (@genes) {
        local $\ = "\n";                 # adds newline to print
        print join "\t", $reg, $stat, $_;
    }
}

__DATA__
region_id   statistic   genelist
1   2.5 A, B, C
2   0.5 B, C, D, E
3   3.2 <NA>
4   0.1 E, F

Output:

region_id       statistic       genelist
1       2.5     A
1       2.5     B
1       2.5     C
2       0.5     B
2       0.5     C
2       0.5     D
2       0.5     E
3       3.2     <NA>
4       0.1     E
4       0.1     F

Answer 3

There are a few ways to do it. This way works, although there may be better ways...

library(stringr) # for str_split
join(subset(a, select=c("region_id", "statistic")), 
     ddply(a, .(region_id), summarise, gene=str_split(genelist, ",\\S*")[[1]]))

Needs plyr and stringr loaded.

Oh, here's a better way:

ddply(a, .(region_id), 
      function(x) data.frame(gene=str_split(x$genelist, ",\\S*")[[1]], 
                             statistic=x$statistic))

Answer 4

Here is a way to do it without any libraries:

data<-cbind(region_id=1:4, statistic=c(2.5, 0.5, 3.2, 0.1), genelist=c("A, B, C", "B, C, D, E", NA, "E, F"))

do.call(rbind, 
        apply(data, 1, 
              function(r) do.call(expand.grid, 
                                  c(unlist(r[-3]), 
                                    strsplit(r[3], ", ")))))

Output:

      region_id statistic genelist
1          1       2.5        A
2          1       2.5        B
3          1       2.5        C
4          2       0.5        B
5          2       0.5        C
6          2       0.5        D
7          2       0.5        E
8          3       3.2     <NA>
9          4       0.1        E
10         4       0.1        F

Answer 5

这是另一个使用plyr

ddply(a, .(region_id), transform, gene = str_split(genelist, ',')[[1]])

Answer 6

A Perl solution:

#!/usr/bin/perl
<>;
print "region_id\tstatistic\tgene\n";
while(<>) {
  chomp;
  my ($reg, $stat, $genes) = split /\s+/, $_, 3;
  foreach my $gene (split /,\s*/, $genes) {
     print "$reg\t$stat\t$gene\n";
  }
}

Just pipe the original file through this script into the output file.

Currently the output values are tab-seperated and not right-flushed, but you can fix that if it is really needed.