Clustering approach in DNA microarray analysis

Microarray DNA technology has paved the way for investigators to expressed thousands of genes in a short time. Analysis of this big amount of raw data includes normalization, clustering and classification. The present study surveys the application of clustering technique in microarray DNA analysis.

We analyzed data of Van’t Veer et al study dealing with BRCA1 and BRCA2 mutations in breast cancer. It was consisted of 18 patients with BRCA1 and 2 patients with BRCA2 mutation. Gene expression data were clustered using hierarchical and non-hierarchical approach. Then different clustering approaches were compared according to the actual classification with R software.

Hierarchical clustering showed a sensitivity of 94% and specificity of 100% in detecting BRCA1 gene. These figures were 89% and 100% for non-hierarchical clustering, respectively, indicating a satisfactory performance for both approaches. All clustering approaches classified sample No. 95 in BRCA2 group, however, clinical manifestations put her in BRCA1 group.

With respect to satisfactory coincidence between clustering and actual classification results, clustering approach could be applied for cases when actual classification is missing.