Determination of signature genes of breast cancer subtypes evaluated by clinical criteria using biclustering algorithm

Several classic clustering techniques have been used to analyze gene expression profiles throughout the last few years. The reliable identification of breast cancer subtypes, on the other hand, is a challenge. In this study, a modified BICLIC biclustering algorithm, termed DBSCAN- BICLIC, was proposed to identify signature genes of breast cancer subtypes. DBSCAN- BICLIC provides the optimized strategy to cluster seeds accurately. Clinical information was used to evaluate experimental results on 2509 breast cancer conditions. The evaluation caused grouping breast cancer conditions into 22 groups; each has different clinical information. Conditions with the same clinical criteria will be grouped together. One interesting aspect of the experiment is that, DBSCAN- BICLIC found the biomarkers of each group of conditions. As a result, each group of similar breast cancer conditions has its signature genes. DBSCAN- BICLIC algorithm has been run for 10 groups. In this study, we present the top five effective signature genes that have been generated for each group. As a total, DBSCAN- BICLIC has generated 40 signature genes in all 10 groups. 32 genes of them have been validated as signature genes according to the literature. The 32 genes are highly effective prognostic genes of breast cancer. This study could contribute in the targeted therapy, and personalized medicine.