The above scheme showed a simple length-dependent weighting for gaps. Thus two isolated gaps give the same score as two consecutive gaps. It is possible to generalise the algorithm to allow gaps of length greater than 1 to carry weights other than the simple sum of single gap weights [27]. Such gap weighting can give a more biologically meaningful model of transitions from one sequence to another since insertions and deletions of more than one residue are not uncommon events between homologous protein sequences. Most computer programs that implement dynamic programming allow gaps to be weighted with the form where is the gap length and and are constants , since this can be computed efficiently [28].

geoff.barton@ox.ac.uk