Computational approaches to dissect ovarian cancer heterogeneity

The project aims to characterize stage I Ovarian Cancer histotypes from a genomic and transcriptomic point of view, integrating different sources of omics data in order to understand their interplay and their potential role in carcinogenesis. In particular, as a first step using a large and multicentric cohort we explored the landscape and the variability of CNV profiles among patients with stage I EOC. The results are reported and published in the first article of this collection, published in the European Journal of Cancer. We define three common genomic instability patterns, namely stable, unstable, and highly unstable based on the percentage of the genome affected by Somatic Copy Number Alterations (SCNAs) and on their length. The genomic instability pattern was strongly predictive of patients’ prognosis. Our next goal was to explore some of the mechanisms that cause the three major patterns of copy number instability. For this purpose, we measured the activity of copy number pan-cancer signatures across the samples and we assessed how these activity levels distinguished the samples. The results of this work are reported in the second article of this collection, and will be submitted soon to a scientific journal. We identified four sample clusters related to different activities among the signatures and one of the clusters (GA2) resulted to be associated with progression-free survival (PFS). As a third step we decided to add another layer of information by adding miRNA expression and TF expression data for the same cohort, starting our process of Multi Omics integration. In this direction, we explored for the first time the expression of isomiRs across different Stage I epithelial ovarian cancer (EOC) histotypes, in order to shed new light on their biological role in tumor growth and progression. Our results are reported in the third article of this collection, and were published in the International Journal of Cancer. We were able to identify a total of 42 histotype biomarkers for all the five subtypes. Using integrative models, we found that the 38% of miRNA expression alterations is the result of copy number variations while the 17% of differential transcriptional activities. Since Multi Omic integration is the keyword of the project, the final step of this work is dedicated to the development of gINTomics, a new R package for Multi Omics integration. gINTomics is currently available on GitHub and will be also submitted to Bioconductor. It is designed to detect the association between the expression of a target and of its regulators, taking into account also their genomics modifications such as Copy Number Variations (CNV) and methylation. With the work performed during my PhD project, we were able to provide new insight on the molecular features of stage I Ovarian Cancer samples from a transcriptomic and genomic point of view, paving the way for a better understanding of the early stage of the disease. The work has both a diagnostic and prognostic potential given the identification of molecular mechanisms and biomarkers. Finally the gINTomics R package represents an interesting resource for the scientific community, giving the possibility to integrate different omics and to better understand their interplay. ​ ​