Re: How can I count the number of substitutions of one letter in a string by another letter in the other string?

[Eliya]

in reply to How can I count the number of substitutions of one letter in a string by another letter in the other string?

In case performance matters and your sequences are longer, you could also use XOR bit logic in combination with tr///:

#!/usr/bin/perl -w
use strict;

my @cnt = qw(A2G G2A C2T T2C A2T A2C G2T G2C);
my %cnt;

for my $change (@cnt) {
    my ($c1, $c2) = split /2/, $change;
    $c2 =~ tr/ATCG/HRDZ/;
    my $x = $c1 ^ $c2;
    $cnt{$change} = eval "sub { \$_[0] =~ tr/$x/$x/ }";
}

my $seq1 = "AAAGATCGTG";
my $seq2 = "AGTACTTTCC";

(my $tmp = $seq2) =~ tr/ATCG/HRDZ/;
my $diff = $seq1 ^ $tmp;
    
print join("; ", map { $_."=".$cnt{$_}->($diff) } @cnt ), ".";

__END__
A2G=1; G2A=1; C2T=1; T2C=1; A2T=1; A2C=1; G2T=1; G2C=1.

The idea here is that XORing two strings produces a different value X for each combination of characters, which you can then count with a simple tr/X/X/.  The additional prior transformation tr/ATCG/HRDZ/ on one of the strings is required because the XOR operation would otherwise produce symmetric/undirected results, i.e. A2G would be indistinguishable from G2A, etc.

(Just in case it isn't clear:  the %cnt hash doesn't hold the count results, but rather the subs doing the respective tr/X/X/ counting.)

[BrowserUk]

Whilst orders of magnitude faster, as posted, your solution is missing some possibilities.

Given the input of:

$seq1 = 'AAAACCCCGGGGTTTT';
$seq2 = 'ACGTACGTACGTACGT';

It should produce 12 counts:

{  
    A2C => 1,  A2G => 1,  A2T => 1,  C2A => 1,  C2G => 1,  C2T => 1,  
    G2A => 1,  G2C => 1,  G2T => 1,  T2A => 1,  T2C => 1,  T2G => 1,
}

But only produces the following eight:

A2G=1; G2A=1; C2T=1; T2C=1; A2T=1; A2C=1; G2T=1; G2C=1.

I'm sure that this is eminently fixable.

[Eliya]

your solution is missing some possibilities.

Well, I implemented (on purpose) exactly the list the OP asked for.  It can of course be extended, if so desired:

my @cnt = qw(A2G G2A C2T T2C A2T A2C G2T G2C ...);

Alternatively, simply compute the full matrix of count routines:

#!/usr/bin/perl -w
use strict;

my %cnt;
my @c = qw(A T C G);
for my $c1 (@c) {
    for my $c2 (@c) {
        next if $c1 eq $c2;
        (my $t = $c2) =~ tr/ATCG/HRDZ/;
        my $x = $c1 ^ $t;
        $cnt{$c1."2".$c2} = eval "sub { \$_[0] =~ tr/$x/$x/ }";
    }
}

my $seq1 = 'AAAACCCCGGGGTTTT';
my $seq2 = 'ACGTACGTACGTACGT';

(my $tmp = $seq2) =~ tr/ATCG/HRDZ/;
my $diff = $seq1 ^ $tmp;
    
print join("; ", map { $_."=".$cnt{$_}->($diff) } sort keys %cnt ), ".
+";

__END__
A2C=1; A2G=1; A2T=1; C2A=1; C2G=1; C2T=1; G2A=1; G2C=1; G2T=1; T2A=1; 
+T2C=1; T2G=1.