Identification of cancer associated biomarkers by analysing biologically enriched clusters using MSAGK_CL

Recent clustering techniques have witnessed the rise of fuzzification systems designed for various medical applications. Computational tools, in the last decades, have considerably empowered the tasks of physicians and biologists in finding cancer-mediating biomarkers. By taking advantage of these benefits, we introduce a fuzzy clustering-based methodology to find genes associated with particular cancers. The work (MSA ^GK_CL ) applies the Gustafson-Kessel (GK) algorithm to gene expression datasets, including normal and carcinogenic states, showing its effectiveness on gene expression profile. The adaptive behaviour of the GK algorithm provides a robust treatment of non-spherical cluster shapes, which is crucial in gene expression analysis. To overcome the problem of deciding the number of clusters optimally, we used cluster validity indices like Xie-Beni (XB), Fukuyama-Sugeno (FS), and Dunn Index. These indices give a quantitative basis for the quality of clustering and help in deciding the best clustering configuration. The second approach was a dynamic threshold-based membership score analysis to identify significant genes. This analysis calculates the maximum absolute differences in membership scores between cancerous and normal datasets using the percentile of these differences to adaptively select the threshold. Its effectiveness is validated by the study's findings through precision, recall, and F1-score metrics, demonstrating its usefulness in discovering genes related to cancer. Such an integrative framework thereby opens up a promising path for future research in the area of cancer genomics, helping identify therapeutic targets for personalized medicine