Individual records from the coding of molecular
polymorphism (molecular profiles) are particularly
useful for the identification of clones or cultivars,
in pedigree analysis, in the estimation of genetic
distances and relatedness, and as a tool in genome mapping
and population genetics. A parametric
statistical analysis of molecular profile components
can be infeasible because of the huge number
of observed markers, the presence of missing values and
the high number of parameters required
to evaluate the importance of interactions among markers.
Moreover, new powerful molecular
techniques make possible the analysis of numerous markers
at one time; therefore parametric
statistical methods could result in troublesome models with
more parameters than data. The field
of computer-based techniques offers new strategies to cope
with the complexity of molecular
profiles. We suggest the use of a Genetic Classifier
System to evaluate the importance of profile
components. The procedure is based on a Genetic Algorithm
approach, a numerical technique that
simulates some features of the natural selection process
to solve problems. A set of isozyme data
from a Norway spruce population is analysed in order to
assess their ability to predict the
individual plant response to the presence of abiotic
stresses. The results, obtained by three different
computer simulations, show that this computer-based approach
is particularly effective for ranking
profile components according to their relevance. Genetic
Classifier Systems could also be used as a
preliminary step to reduce the complexity of molecular data sets.