(732b) Integrative -Omics Analysis of Cancer Protein Secretion

Early detection and diagnosis of cancer is critical to the successful treatment of the disease, as later events such as metastasis can severely reduce the chances of survival. Although a number of screening methods exist for different cancers, many require an invasive surgical procedure or suffer from poor sensitivity and/or specificity. This has motivated efforts to identify biomarkers present in biofluids (e.g., blood or urine) that could accurately predict the presence or progress of cancer within an individual without the need for more invasive approaches. However, the heterogeneity of the disease has made such efforts nontrivial, exhibiting substantial variation across different individuals and tissues-of-origin, and even among cells comprising the same tumor. A systems-level approach is therefore necessary to account for the governing biological networks contributing to such behavior, and to identify core patterns for the development of highly specific cancer biomarkers.

To identify potential biomarkers likely to be detectable in patient biofluids, we focused specifically on the subset of proteins processed by the secretory pathway, termed the “secretome.” Secretome members were selected based on their annotation in the Human Protein Atlas (HPA), and the presence of an N-terminal signal peptide. Transcriptomic (RNA-Seq) data from solid tumors and matched normal tissues spanning over 20 different cancer types were retrieved from The Cancer Genome Atlas (TCGA), and analyzed using a combination of statistical and integrative approaches to predict candidate biomarkers. A differential expression analysis was performed for each individual cancer type between tumor and matched normal tissue, to reduce tissue-specific effects unrelated to cancer. These results were combined with absolute transcript abundance profiles spanning all normal tissue samples, and cross-referenced with the human plasma proteome to identify a subgroup of upregulated secretome genes as potential biomarkers for each cancer type. The resulting collection of candidate biomarkers included a number of previously-identified proteins known to be involved in malignant transformation or tumor development, as well as many new candidates whose role in cancer had not yet been explored. Beyond providing a collection of potential biomarkers to improve cancer detection and diagnostics, we elucidated cancer-specific and pan-cancer patterns in candidate protein features, such as combinations of post translational modifications, and their connectivity in protein-protein interaction networks. These results provide a deeper understanding of the cancer-specific protein secretory program, which is a critical step toward improving the detection and treatment of the disease.