On 2013 Oct 23, Albert Erives commented:

This paper describes an example of what I have been calling an equivalence class of transcriptional enhancers. This is compared with other such sets in this short collection of slides: How to best approach transcriptional enhancers?. Albert Erives. figshare. http://dx.doi.org/10.6084/m9.figshare.829049

What is not well-described in this paper is how their common site organization was found based on a subset of "specialized sites" within homotypic site clusters. A specialized site is defined mathematically as a sequence matching a binding motif descriptor that identifies a single unique site within each member of a set of homotypic binding site clusters. In other words, specialized sites are identifiable and distinguishable by the existence of any motif that matches them as single components of each homotypic site cluster belonging to a set of mechanistically-equivalent enhancers. A motif that describes more than a single site in each cluster is therefore not a specialized motif. Okay, that's not the easiest thing to get across even today, but this method led to the identification of the so-called D-alpha site, which turned out to be a Su(H) binding site that had evolved over a relic Dorsal binding site, and the so-called D-beta site, which turned out to be the closest Dorsal-binding-like site to the Twist site, the distance to which we later found to encode a large component of the response to the Dorsal morphogen gradient. (See here Crocker J, 2008 and here Crocker J, 2010.) For me, the major significance of this method is that it works well in distinguishing functional sites from relic (pseudogenized), fast-evolving sequences.

On 2013 Oct 23, Albert Erives commented:

This paper describes an example of what I have been calling an equivalence class of transcriptional enhancers. This is compared with other such sets in this short collection of slides: How to best approach transcriptional enhancers?. Albert Erives. figshare. http://dx.doi.org/10.6084/m9.figshare.829049

What is not well-described in this paper is how their common site organization was found based on a subset of "specialized sites" within homotypic site clusters. A specialized site is defined mathematically as a sequence matching a binding motif descriptor that identifies a single unique site within each member of a set of homotypic binding site clusters. In other words, specialized sites are identifiable and distinguishable by the existence of any motif that matches them as single components of each homotypic site cluster belonging to a set of mechanistically-equivalent enhancers. A motif that describes more than a single site in each cluster is therefore not a specialized motif. Okay, that's not the easiest thing to get across even today, but this method led to the identification of the so-called D-alpha site, which turned out to be a Su(H) binding site that had evolved over a relic Dorsal binding site, and the so-called D-beta site, which turned out to be the closest Dorsal-binding-like site to the Twist site, the distance to which we later found to encode a large component of the response to the Dorsal morphogen gradient. (See here Crocker J, 2008 and here Crocker J, 2010.) For me, the major significance of this method is that it works well in distinguishing functional sites from relic (pseudogenized), fast-evolving sequences.