promoterwise is a sort of next generation DBA (see next section). It is designed for comparisons between two promoter sequences or realistically any two orthologous regulatory regions (or homologous for that matter, but in theory it should work better for orthologous regulatory regions, depending on how much active change you expect paralogous regulatory regions to have). Promoterwise reports alignments between these two sequences assumming that alignments cannot overlap in both sequences, but *not* assumming that the alignments have to be co linear or on the same strand.
Promoterwise works by taking the two sequences and then finds all common exact 7mers between them, in both the forward and reverse strands. These are then merged such that close HSPs (whoes centers are within the window size of each other) are considered one region. These regions then have a local version of the DBA algorithm run over them, which has a model of DNA similarity of small regions of similarity, potentially with small gaps separated by large pieces of unknown DNA.
The resulting set of alignments are then sorted by score, and a simple greedy algorithm is used to discard ``bad'' subsequent alignments. By default this is to discard alignments which overlap on the query coordinate with alignments of a higher score (this can be changed). The alignments are then outputted with bits score. In my hands I think a bit score of over 20bits looks good.
Of course there are many options to change here.