GENETIC ALGORITHM FOR APPLYING CONSTRAINTS IN CHROMOSOME CLASSIFICATION

A genetic algorithm was used to solve the optimization problem of jointly classifying human chromosomes subject to constraints imposed by the context of a metaphase cell. When applying the constraint of no more than two chromosomes per class to two widely used data sets, error rates as low as in the previously best-known classification method were obtained, confirming that a genetic algorithm is indeed a suitable tool for investigating this problem. Applying a further chnstraint based on the presumed close similarity of the banding pattern of homologous chromosome pairs resulted in only a small improvement in error rates, but did assist in the discrimination of chromosomes that should be rejected.