Actually there's another approach to this, not using necessarily using Genetic algorithms or regex - weight matrices. So you would want to look through a set of data, basically scoring it for the occurance of pattern(s). As you find the common pattern, you create a weight matrix, based upon the number of occurances and the frequency of occurance of a particular letter/resiidue at a position in the pattern. Once you have some examples of the pattern you can rescan your new data for the occurance of the pattern, reinforce the matrix by rescoring it for new data, etc. Check out this link: TFBS. Boris lenhard of the Karolinska Institute presented this as a way of scanning for promoter sequence patterns in July. It should be available for download.

MadraghRua
yet another biologist hacking perl....