Figure 10

Counterintuitive example for simple non-pairwise indel scoring. An example showing why the accuracy of alignments might decrease with increasing the reticular threshold when the gaps are scored with simple non-pairwise indel scoring. a) The best alignment found with t = 0 threshold and simple non-pairwise indel scoring. Only the locally optimal alignments were kept in the progressive alignment, thus the final multiple alignment contains a few, aggregated gaps, for which the pairwise alignments are also optimal. b) The best alignment found with t = 800 threshold value and simple non-pairwise indel scoring. This alignment scored better than the previous alignment, because it contains more homologous pairs. The increased number of homologous amino acid pairs is achieved by inserting more gaps, however, the increased number of gaps do not reduce too much the score as simple non-pairwise indel scoring is applied. ClustalW never considers this alignment as it can be built only via sub-optimal alignments.