75. Copy number variation heterogeneity as the measure for biological consistency in hierarchical cancer classifications

Cancers are heterogeneous diseases with unifying features of abnormal and consuming cell growth, where the deregulation of normal cellular functions is initiated by the accumulation of genomic mutations in cells of - potentially - any organ. At diagnosis malignant tumors present with patterns of somatic genome variants on diverse levels of heterogeneity. Among the different types of genomic alterations, copy number variants (CNV) represent a distinct, near-ubiquitous class of structural variants. Cancer classifications such as the National Cancer Institute Thesaurus (NCIt) provide large sets of hierarchical cancer classification vocabularies and promote data interoperability and ontology-driven computational analysis.

However, high heterogeneity in cellular phenotypes and dynamic plasticity of tumor microenvironments make tumor categorization a demanding and complicated task with the need to balance between categorical classifications and individual, 'personalized' feature definitions. To find out how categorical classifications reflect biological facts, we conducted a meta-analysis of inter-sample genomic heterogeneity at different levels of the classification hierarchies based on genome-spanning CNV profiles from 97,142 individual samples across 512 hierarchical cancer entities in the progenetix database. The use of a large data set of individual cancer samples allows for a greater exploration of genomic tumor heterogeneity between and inside given diagnostic concepts. With this study, we applied hierarchical clustering to quantify the heterogeneity among cancer entities through a refined measure of hamming dissimilarity based on CNV events. The results point out common/specific CNV patterns and potential subtypes of cancer entities, which will help in the improvement of patient stratification and current cancer classification.