Meta–analysis of genomic and expression data in endometrial cancer

Abstract

Endometrial carcinoma is the commonest gynaecological cancer in the world, with huge hormonal influence, as it occurs in the inner lining of the uterus where the cell proliferation, regeneration, and functions are maintained through hormonal influence. But there is a lack of researches done on the analysis of the genetic level hormonal influence of endometrial cancer. Mostly endometrial carcinoma is diagnosed when the carcinoma is confined to the uterus and also current therapeutics fail to treat late-stage disease. This study was conducted to identify meaningful subtypes in endometrial carcinoma and to identify the hormonal influence in each subtype. Study will help in improving the efficiency of the treatments and reducing the toxicity of the treatments by identifying clues to find target therapeutics through the analysis of genetic level hormonal influence. Gene expression data obtained from cBioportal has been analysed in this research using unsupervised learning techniques since gene expression data illustrate the biological processes happening inside the cells. Partitioning around medoids, K-means and Hierarchical clustering techniques used in initial clustering and hierarchical clustering technique has been used with different distance and linkage measures, to identify meaningful clusters. After the cluster identification, cluster validation has been conducted according to internal measures like Silhouette, Dunn index and relative measures to identify optimal number of clusters. External validations like comparing the classes with clinical variables and visual analysis of the classes using heatmaps has also been conducted. Identified clusters are analysed to find the hormonal influence inside each cluster and stage-specific hormonal influence also analysed using feature filtering methods like Between Sum of Square / Within Sum of Square. After heatmap filtering it was identified three cluster analysis results have meaningful clusters, out of them identified the cluster analysis with 15 clusters from hormonal gene dataset having significant genes in each cluster.