Gene Discovery in Leukemia Revisited: A Computational Intelligence Perspective

An approach using clustering in combination with Rough Sets and neural networks was investigated for the purpose of gene discovery using leukemia data. A small number of genes with high discrimination power were found, some of which were not previously reported. It was found that subtle differences between very similar genes belonging to the same cluster, as well as the number of clusters constructed, affect the discovery of relevant genes. Good results were obtained with no preprocessing applied to the data.Nous avons \ue9tudi\ue9 une approche faisant appel au groupage, en combinaison avec les ensembles approximatifs et les r\ue9seaux neuronaux, pour la d\ue9couverte des g\ue8nes au moyen de donn\ue9es sur la leuc\ue9mie. Nous avons d\ue9couvert un nombre r\ue9duit de g\ue8nes poss\ue9dant un pouvoir discriminant \ue9lev\ue9, dont certains n'avaient pas \ue9t\ue9 r\ue9pertori\ue9s ant\ue9rieurement. Nous avons constat\ue9 que des diff\ue9rences subtiles entre des g\ue8nes tr\ue8s semblables appartenant \ue0 la m\ueame famille, ainsi que le nombre de familles construites, avaient un impact sur la d\ue9couverte des g\ue8nes pertinents. Nous avons obtenu de bons r\ue9sultats, sans pr\ue9traitement des donn\ue9es.NRC publication: Ye