Edit-tree coding of similar sequence groups. Circles represent DNA sequences in a database; the straight-line distance between circles represents the edit distance between sequences. Initially (a) we are presented with the input database. In the first step (b), groups of similar sequences are discovered. In the second step (c), each group is edit-tree coded independently by determining a reasonable tree, selecting a root sequence (coloured black) and recording the necessary edits along each edge. Some sequences are not sufficiently similar to any other sequence to be delta-encoded – these sequences will be recorded verbatim.