How about a hash with col* as keys and data_line* as an array stored for the corresponding key?
Well lets see what to do...

my %database;
my $curCol;
for(<DATASET>)
{
  if(/^(col\d+)\s+(.*)/)  #(.+) isnt good but as I dont know exactlly 
+how your data looks like...
  {
    push(@{$database{$1}}, $2);
    $curCol = $1;
  }
  elsif(/^\s+(.*)/)
  {
    push(@{$database{$curCol}}, $1);
  }
}
[download]

Maybe I also expalin a little bit, but I didnt try out the code so there might be some bugs in it, I didnt notice. You create a hash of arrays with col as key and push each line that has no new col into the previous col. If there is a new col a new key and array is initiated and so on. I hope it is logical and helpfull.