The advance. Last year, Joshua Denny and colleagues at Vanderbilt University published the first study that demonstrates the feasibility of associating genetic modifications with data on phenotypic traits mined from electronic medical records1. The approach, which they called PheWAS (for phenome-wide association scans), is akin to the genome-wide association studies (GWAS) widely used today to find single-nucleotide polymorphisms (SNPs) that are genetically linked in a population to a particular disease trait—except that PheWAS is GWAS in reverse. GWAS associates genotypes with a given phenotype, such as height or a genetic disease. In contrast, PheWAS attempts to determine the range of clinical phenotypes associated with a given genotype.
The Vanderbilt group analyzed the medical records of ∼6,000 patients who had been tested to see whether they carried a total of five SNPs previously associated with seven diseases (coronary artery disease, carotid artery stenosis, atrial fibrillation, multiple sclerosis, lupus, rheumatoid arthritis and Crohn's disease). To identify patient phenotypes in an automated fashion, they used billing codes in the electronic medical records to group patients into 'case' and 'control' populations for 776 phenotypes. Finally, a Chi-squared statistical test was used to evaluate whether patients harboring a specific SNP also tended to display a particular phenotype. The authors noted that, although there are many statistical challenges with this kind of analysis and there is much room for improvement of their method, four of the seven previously known disease-gene associations could be replicated, and several potential associations with other diseases were identified but not rigorously validated. These results highlight the possibility that novel biological discoveries might be made using this approach.
This is a preview of subscription content, access via your institution