[kman42]

I'm trying to do a little bioinformatics and I'm sure there is a simple way to do what I want using unix utilities that I'm just not aware of. Perhaps grep or something similar. Basically, I want to take a text sequence (dna) where some of the letters are variable and try to match a six-letter string to it.

Here's an example:

TG[T C]TG[T C]TG[T C]GG[T C A G]GC[T C A G]ATGCC[T C A G]CA[A G][C A]G[T C A G]

Letters in brackets can be any one of those letters. I'd then like to search for matches on this string using the following strings:

GGATCC
CCTAGG
GAATTC
CTTAAG

I'm sure this is fairly easy using one of the builtin unix string utilities, but I'm at a loss. Can someone help me out?

thanks,
kman

[kman42]

Originally posted by kman42:
I'm trying to do a little bioinformatics and I'm sure there is a simple way to do what I want using unix utilities that I'm just not aware of. Perhaps grep or something similar. Basically, I want to take a text sequence (dna) where some of the letters are variable and try to match a six-letter string to it.

Here's an example:

TG[T C]TG[T C]TG[T C]GG[T C A G]GC[T C A G]ATGCC[T C A G]CA[A G][C A]G[T C A G]

Letters in brackets can be any one of those letters. I'd then like to search for matches on this string using the following strings:

GGATCC
CCTAGG
GAATTC
CTTAAG

I'm sure this is fairly easy using one of the builtin unix string utilities, but I'm at a loss. Can someone help me out?

thanks,
kman

I tried typing in the long sequence as a regular expression and saving as a file and then using grep to search for the short sequences, but that doesn't work. Can you only use regular expressions in the pattern, not the file?

kman

[kman42]

Perhaps the perl m// function?

kman

[indra]

this little perl example may help:

#!/usr/bin/perl

$_ = 'TGTCTGTCTGTCGGTCAGGCTCAGATGCCTCAGCAAGCAGTCAG';

$pat1 = 'GGATCC';
$pat2 = 'CCTAGG';
$pat3 = 'GAATTC';
$pat4 = 'CTTAAG';

/($pat1)|($pat2)|($pat3)|($pat4)/ && print "matched $1\n";

[Camelot]

If the source is one long DNA string (and not just the 20 or so characters you describe), neither grep nor perl is going to do this for you.

The reason is that both models work at the line level - read a line, check for match, move onto next line.

If you have one long DNS sequence, it will read in the entire sequence as one 'line' and report if there's any match - anywhere in the string.

Assuming you can read the entire sequence into RAM, there's likely to be at least one match somewhere in the string, but both models will simply tell you there's a match, not where it is, nor if there's more than one match in the sequence.

You'd have to change the code to something that reads a few bytes at a time, checks for matches, moves along a byte, check for more matches, etc, etc. ad infinitum.

In short, there's a reason why there are commercial gene sequencing programs costing $$$$$s.