Amid the buzz of the results from the Cancer Genome Atlas and a multitude of studies looking at the epigenomics of cancer at this year's annual meeting of the American Association for Cancer Research, scientists discussed the future of genome-wide association studies. How should scientists take these large data sets and move them forward to translate them into information that is useful for both the diagnosis and treatment of cancer?
At a forum addressing the future of GWAS in personalized medicine, Harvard's David Hunter began by talking about the process of performing a genome-wide association study and the increasingly important role of replication studies. In the discovery phase, the challenge is wrapping your head around all that data, especially if you're not used to a systems biology approach, Hunter said. "Naturally [for] someone who's used to seeing data based on a single SNP, the first challenge is how [to] integrate the data [and] how [to] present it," he said. For the replication phase, Hunter sees a growing need for larger-sized studies to effectively find all the common variants with largest effects in cancer, a process that he thinks scientists should aim to complete in the next couple of years. Despite the concern that association studies are a dead end, he thinks that it's "rational, sensible, and cost-effective to keep doing GWAS while we await the cost-effectiveness of whole genome sequencing."
In the past several years, the number of identified disease susceptibility loci for various cancers has skyrocketed. Culling data from the 2007 AACR meeting, Hunter said that prostate, breast, and colon cancer each had one common variant associated with them; two years later, prostate cancer weighed in with at least 18 variants while breast and colon cancers boast a whopping 40 or more variants each. With that in mind, Hunter believes GWAS will continue to play a serious role in mapping the etiology of cancer, searching for gene-specific mechanisms of diseases, and finding out more about the role of intergenic regions, since many association loci are found within noncoding regions.
While GWAS have increased the number of available risk factors, current risk prediction algorithms are still not reliable enough for clinical use. While Hunter believes that one-time screening for multiple lower penetrance conditions will be part of the clinical future, he also emphasized that scientists must complete the discovery phase for common alleles in common diseases in order to move forward on improving risk prediction. Until the current universe of risk variants is expanded to include all variants, he noted, scientists, physicians, and direct-to-consumer genotyping companies should hold off on considering the information clinically useful.