In this article, we compare the performance of 19 cluster validity indices, in identifying some possible genes mediating certain cancers, based on gene expression data. For the purpose of this comparison, we have developed a method. The proposed method involves cluster generation, selection of the best k-value or c-values, cluster identification, identifying the altered gene cluster, scoring an altered gene cluster and determining the best k-value or c-value exploring through biological repositories. The effectiveness of the method has been demonstrated on three gene expression data sets dealing with human lung cancer, colon cancer, and leukemia. Here, we have used three clustering algorithms, i.e., k-means, PAM and fuzzy c-means. We have used biochemical pathways related to these cancers and p-value statistics for validating the study.