In contrast, in the teleost HoxA cluster example the numerical score of the anchored alignment was around 15% above the score of the non-anchored alignment. This demonstrates that the greedy optimisation algorithm used by DIALIGN can lead to results with scores far below the optimal alignment. In such situations, improved optimisation algorithms may lead not only to mathematically higher-scoring alignments but also to alignments that are closer to the biologically correct alignment. We will use our anchored-alignment approach systematically to study the efficiency of objective functions and optimisation algorithms for our segment-based approach to multiple sequence alignment.