Scaling up for a deeper look at diabetes
Image by Nadav Kupiec, Broad Communications
The current era of high-throughput genome scanning brings with it a surge in the discovery of genetic markers that confer risk to human disease. These findings result from whole genome association studies, in which the genomes of many people are scanned on microarrays to locate genes related to diseases like type 2 diabetes, multiple sclerosis, and Crohn’s disease. From these scans, researchers have identified the critical “first fruits” of association studies — the genetic changes with the greatest effects on disease risk. But most common diseases are impacted not by one or even a few genetic changes, but by a large number of them, each with relatively small contributions.
Finding the additional genes with slighter impacts is a real challenge. Most association studies to date have involved a few thousand people, well below what is needed to thoroughly prove such genes’ connections to disease. Accomplishing that requires expanded studies with much larger sample sizes, providing a boost in statistical power that can help detect modest genetic effects.
One of the first such expansions has recently been completed by researchers at the Broad Institute of MIT and Harvard, along with collaborators in the Wellcome Trust Case Control Consortium/U.K. Type 2 Diabetes Genetics Consortium (WTCCC/UKT2D) and the Finland-United States Investigation of NIDDM Genetics (FUSION). In the new work, described in the March 30 issue of Nature Genetics and led at the Broad Institute by Program in Medical and Population Genetics director David Altshuler, data from association studies completed last year by the three groups were combined into a new “meta-analysis” of type 2 diabetes.
The current work discovered six new single-letter differences in DNA, known as SNPs, that increase diabetes risk. Along with the eight risk factors identified in the original work, three of which were novel, the six new variants add to the growing number of DNA changes connected to this disease. The finding also shows promise for this kind of large meta-analysis study to be instrumental in identifying SNPs with subtle effects on common diseases.
The Broad, WTCCC/UKT2D, and FUSION groups had collaborated on their original studies, sharing data before selecting candidate SNPs to follow up in the second phase of the study. “It was our combined evidence that led us to those first three novel genes,” said Richa Saxena, co-first author and postdoctoral researcher in the Broad Institute’s Program in Medical and Population Genetics. “But this time, we really worked together with the other groups as a core analysis team from the beginning.”
By combining efforts, the researchers were able to expand the study in two critical ways. Compared to the few thousand people studied in each of the original scans, the new work combined genetic data on over 10,000 people, both diabetic patients and healthy controls, giving the team more statistical power to find new variants. In addition, the researchers were able to use knowledge generated by the HapMap project to increase the number of SNPs analyzed, from under 400,000 in each original study to roughly 2.2 million in the new study.
DNA is inherited in chunks, and based on those patterns of inheritance — revealed by the HapMap — the scientists were able to use a process of “imputation” to predict the identity of SNPs that were not directly genotyped. “We don’t actually test many of those SNPs, but through imputation, we can guess at them quite reliably if we know how they are linked in the HapMap,” said Ben Voight, co-first author on the new study and a postdoctoral researcher in the Broad Institute’s Program in Population and Medical Genetics.
Of the 2.2 million SNPs in the study’s first phase, 69 promising SNPs were re-tested in a replication stage of genotyping in over 22,000 people. Of those, eleven SNPs were different enough among diabetics and healthy controls to be studied in a third stage of over 57,000 samples from a total of 13 research groups comprising the Diabetes Genetics Replication and Meta-Analysis (DIAGRAM) Consortium. In that final stage, six of the candidate SNPs were reliably linked to diabetes.
The new SNPs serve as useful guideposts, representing a region of DNA that may contain the true causal variant. But in order to determine the exact genetic changes that confer risk, the researchers must conduct fine-scale analyses, including sequencing the DNA surrounding the identified SNPs. Once found, these DNA changes must then be tested in the laboratory in cellular and animal models, which may give insights into the biology of diabetes. “Doing this kind of unbiased association study points you to new genes that you would have never considered,” said Saxena. “They open the door to understanding how genes influence disease, but you still must do the biological work to discover how that happens.”
In addition to studying the biology of these SNPs, the consortium may join forces with additional groups in the future to conduct an even larger scale meta-analysis, with the goal of unraveling the increasingly complex nature of this common disease. Comparing type 2 diabetes to a jigsaw puzzle, Voight explained, “We can’t yet envision the whole picture, but we’re starting to gather the pieces.”
Other Broad researchers contributing to the work include Paul de Bakker, Kristin Ardlie, Noel Burtt, Mark Daly, Lauren Gianniny, Candace Guiducci, and Finny Kuruvilla.