I have a script which extracts equal length DNA-sequences at user-specified positions from a genome and then calculates the frequency at which each letter (A/G/C/T) occurred at each position.

The user-specific positions can also come with a frequency - how many times that point was seen.

If this frequency is ignored (i.e. position-weighted analysis), the script runs at a satisfactory speed. However, it is when the frequencies are taken into consideration that it is considerably slow.

My approach for this has been:

if ($mode eq "Position") {
    push (@data, $string);
} elsif ($mode eq "Freq") {
    push (@data, $string) for (1..$F[2]);
}
    
    my %freq;
for my $i (1..@data) {
    $freq{$1}[$-[0]] += 1/@data while $data[$i-1] =~ /(.)/g;
}
[download]

Which is evidently inefficient and quite literal - i.e. add that sequence X(freq) number of times into the array.

Is there a better way to do weighted frequency calculations and if so, how?

EDIT: The sequences can range from 30 to 250 characters. Weighting occurs at an average of 26, but with a range of 5-3565 for one particular sample. In any given run, around 200,000 unique positions exist within the input data.