Cluster the pairwise identity matrix

[rocreguant]

It may be useful to offer a ranking scheme that deliberately clusters sequences according to their local similarity to each other (ie maximising the clarity of these patterns), rather than comparing everything to the first sequence.