The recently introduced quality criteria for optimisation, describing the c
overage of the search and solution spaces as well as the reproducibility of
both, are applied in combination with experimental design to fine-tune par
ameter settings and fitness function of a genetic algorithm for the structu
re optimisation of a heptapeptide. A series of influences of the investigat
ed parameters are revealed by these criteria, while none of them seem signi
ficant from the fitness values of the last population alone. It is therefor
e suggested to apply these criteria, which are not based on the fitness val
ue of the final population, when developing genetic algorithms. It is shown
that they are easily adaptable to specific problems. (C) 1999 Elsevier Sci
ence B.V. All rights reserved.