TCGA Kidney Cell Atlas: A differential gene expression database for the dissection of tumor-specific gene modules.
e17078 Background: The Cancer Genome Atlas project has become a leading source for data that has allowed the identification of a broad range of human cancer tumor types and subtypes and has revealed deep complexity with respect to the differentiation, or lack thereof, among human cancers. In particular, differential gene expression analyses have revealed a wealth of active oncogenic pathways, underlying gene mutation drivers, discriminative markers, and candidate therapeutic targets. Despite its rich composition, several factors have led to it not attaining the utility it would seem to offer. Methods: To study this, we dissected molecular subtypes in the TCGA and used the Pan-Kidney (n = 1022 samples) Portion within it to determine where obstacles seem to limit its utility. We re-clustered the renal carcinomas to create more appropriate histology annotations for these samples. The molecular subtypes were then found through K-means using differentially expressed known developmental regulators per histological annotation. After deriving these new annotations, the histology and molecular subtypes were compared to one another via T-test to generate gene modules that characterize these classes/subclasses. Results: We identified a number of factors that include inconsistent metadata attributes, apparent misclassification of histological subtypes, and molecular subtypes that do not match with that obtained by focused approaches to rederive principle subclasses. Our gene modules showed a molecular subtype of clear cell renal carcinoma that was enriched for vascular development and nephron development. In general, the clear cell renal carcinoma and papillary renal cell carcinoma cohorts both showed significant co-expression with atlases that were enriched for genes involved in kidney development. Conclusions: Our atlas highlights the limitations of the current TCGA atlas and provides another tool to capture the rich insights from the TCGA repository through the efforts explained above, highlighted by its use in kidney carcinoma.