Breast Cancer Classification and Biomarker Discovery from Microarray Data Using Silhouette Statistics and Genetic Algorithms

Discriminating heterogeneous cancers by microarrays is a topic of much interest in bioinformatics. A number of methods have been proposed and successfully applied to this problem. In this paper, we aim at using genetic algorithms for gene selection and propose silhouette statistics as discriminant function to classify breast cancers for biomarker discovery. Distance metrics and classification rules based on silhouette statistics have also been discussed to improve our algorithms for high classification accuracy. Finally, the proposed method is compared to previously published methods. Many experimental results show that our method is effective to discriminate breast cancer subtypes and find many potential biomarkers to help cancer diagnosis.