One common technique is to only look at all $k$-mers of a sequence: For each input sequence we can store all the consecutive subsequences of length \(k\) occuring in the sequence, and when such a $k$-mer occurs in two sequences, we can then check the corresponding positions in these sequences to see whether they are indeed related. for j in range ( len ( s ) - w + 1 ): window = s [ j : j + w ] subsample = [ window [ Li ] for Li in L ] # The window is identified by the sequence index # and its start position in the sequence. In our approach, we estimate the distance between sequence \(A\) and \(B\) as the average hamming distance between matching windows between sequence \(A\) and \(B\).