[mismith]

For all you CS people who wanted to try your hand at bioinformatics, here's a tiny little problem I am working on. I am an inexperienced programmer, but might enough to fake my way through. I just need a little help getting started.

I will have about 200 strings of 23 letters long, and each letter can be either, A, C, G, or T.

I want to "score" these strings based on a certain letter at a certain positions.

Example string
nnnnXnnnnnnXnnXnnXXXXXnn

In this example I want to add a point for the letter A in the 5th position, and add a point for a T in the 11th position. I also want to subtract a point for a G in the 13th position. All of the n's can be any of the four letters.

I have a website that will search a longer string and pull these patterns out, but I can't figure out a command for this. Does anyone know a simple command to do something like this. I don't really have a language preference, but I know perl is regularly used for DNA sequence analysis.

Thanks.

-MS

[davechen]

Well, I'm a geek and I got sick of my work, so I hacked up a perl script. It should do what you want. The scoring loop is kind of gross. It takes the strings from standard input.

Code:
#! /usr/bin/perl

while(<>) {

# take a 23 character string and make a 23 element array, @letters.  each element
# of @letters is a number, the ascii value of the character.
    @letters = unpack("c23", $_);

    $score = 0;

    for ($i=0; $i<=$#letters; $i++) {

        if (($i==4) && ($letters[4] == 65)) {
# found 'A' in the 5th letter
            $score++;
            print chr($letters[$i]), " ";

        } elsif (($i==10) && ($letters[10] == 84)) {
# found 'T' in the 11th letter
            $score++;
            print chr($letters[$i]), " ";

        } elsif (($i==12) && ($letters[12] == 71)) {
# found 'G' in the 13th letter
            $score--;
            print chr($letters[$i]), " ";

        } else {
# print the non-matching characters in lower case
            print chr($letters[$i]+32), " ";
        }

    }
    print "\n";
    print "score = $score\n";
}

[Mithras]

Also, bear in mind that this is a pretty simplistic framework, if you're trying to find sequences homologous to some ancestral sequence, or bearing some target sequence of interest. Might I suggest you read the first two chapters of Durbin's Biological Sequence Analysis? You can use Amazon's search-inside feature to read the useful bits.

You might look into either
(a) a simple string edit distance measure
or
(b) scoring with log odds ratios against a background model

as discussed in the book.