The authors mapped an entire field of research to find out how certain cancer types are associated with each other.
For this report, the authors built a database of cancer biomarker publications and then categorized and segmented the database into 36 cancer types and 441 molecular entities. [© Mopic - Fotolia.com]
Our goal for this analysis was to ask questions such as:
—How are selected cancer types associated with each other?
—Which molecular entities are "associated" with specific disease types?
—Which combination(s) of molecular entities can provide a unique signature for specific disease types?
—Can the publications dataset predict which signatures are most promising as biomarkers?
In order to answer the questions, we built a broad database of cancer biomarker publications using PubMed data, and then categorized and segmented the database into 36 cancer types and 441 molecular entities.
Highlights of this report:
The ability to map an entire field of research is powerful as it allows us to uncover associations not apparent if we focus in on a niche space.We have utilized a methodology whereby not only do we evaluate the publications record across various cancer classes, but also can probe specific patterns of “molecular entities” as a means to ask whether a particular disease class express this pattern or not—we can then iterate these patterns to evaluate many different combinations of molecular classes.This approach allows us to make specific predictions of patterns, which can subsequently be tested empirically—in this manner, we can produce a short list of potential marker combinations which can then be evaluated on clinically annotated sample collections to validate these predicted patterns.This methodology is scalable and goes beyond the oncology space. We have presented specific cases (focusing primarily on breast cancer) to illustrate the approach; however, we can leverage this approach to other disease classes, and this is crucial as the field of diagnostics and personalized medicine is moving beyond the realm of cancer to include other disease classes. Indeed, we believe that over the course of this decade, the companion diagnostics field will expand beyond the 12 U.S. FDA-approved nucleic acid-based companion diagnostics entities to a much larger number beyond cancer and into cardiovascular disease, CNS disease, infectious disease, metabolic disease, etc. The current pace of deal-making in the companion diagnostics space suggests that some of these partnerships will result in marketed products in 5–7 years downstream.