Research ArticleComputational Medicine

Disease Risk Factors Identified Through Shared Genetic Architecture and Electronic Medical Records

See allHide authors and affiliations

Science Translational Medicine  30 Apr 2014:
Vol. 6, Issue 234, pp. 234ra57
DOI: 10.1126/scitranslmed.3007191

You are currently viewing the editor's summary.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

Medicine by Association

As data get bigger, the challenge is to extract human-sized conclusions that we can comprehend and use. Li and colleagues have done exactly this by exploiting VARIMED, a hand-curated database of single-nucleotide polymorphisms (SNP) associated with diseases or clinical parameters such as cholesterol level and smoking status, extracted from the literature.

By finding pairs of diseases and these nondisease clinical parameters (which they call traits) that are associated with the same SNP variants, they construct hypotheses that the traits could be prognostic markers or risk factors for the disease. Ninety-four of the 120 pairs they identified were known and published in the literature; 26 pairs were previously undescribed. The known associations tended to fall into groups: solid organ cancer with prostate-specific antigen (PSA) and autoimmune disorders with major histocompatibility complex (MHC)–related molecules, for example. The authors were able to validate several of the newly associated traits and diseases by extracting data from electronic medical records from three clinical centers: They found that patients with abnormal mean corpuscular volume were more than three times more likely to receive a diagnosis of acute lymphoblastic leukemia within a year than those with normal values. Similarly, abnormal magnesium levels predicted a greater risk of developing gastric cancer within a year, and abnormally high PSA levels predicted a doubling in the odds of receiving a lung cancer diagnosis within a year.

This all in silico discovery and validation of potential risk factors for disease present an important hypothesis-generating tool for medicine. Prospective clinical trials will test whether these clinical traits can serve as informative diagnostic and prognostic markers.

View Full Text

Stay Connected to Science Translational Medicine