PHILADELPHIA, Pa.—Researchers have discovered asophisticated computational algorithm that, when applied to a large set of genemarkers, has achieved greater accuracy than conventional methods in assessingindividual risk for type 1 diabetes.
A research team led by Dr. Hakon Hakonarson, director of theCenter for Applied Genomics at The Children's Hospital of Philadelphia,suggests that their technique, applied to appropriate complex multigenicdiseases, improves the prospects for personalizing medicine to an individual'sgenetic profile. The study appears in the Oct. 9 issue of the online journal PLoSGenetics.
Genome-wide association studies (GWAS), in which automatedgenotyping tools scan the entire human genome seeking gene variants thatcontribute to disease risk, have yet to fulfill their potential in allowingphysicians to accurately predict a person's individual risk for a disease, andthus guide prevention and treatment strategies.
"For type 1 diabetes (TID), it means that we could identifya high-risk group of individuals who would be followed and monitored closelyfor earliest signs of T1D (before any islet cell destruction) and interventionwould begin with anti-T cell drugs—or other immunosuppressive drugs—to preventT1D from developing," Hakonarson says. "We believe also that development of anew drugs that block the CLEC16A (formerly KIAA0350) signaling pathway in NKcells will become a specific future preventive therapy for T1D."
Hakonarson points out that this approach can also be appliedin other disease areas, with inflammation and autoimmunity most effective.
"We are currently applying this algorithm on other GWAS datawe have and we see marked improvement in disease prediction of otherinflammatory/autoimmune disorders," he says. "This method is likely to workwell on diseases that are highly heritable."
According to Hakonarson, for many diseases, the majority ofcontributory genes remain undiscovered, and studies that make selective use ofa limited number of selected, validated gene variants yield very limitedresults.
"For many of the recent studies, the area under the curve(AUC), a method of measuring the accuracy of risk assessment, amounts to 0.55to 0.60, little better than chance (0.50), and thus falling short of clinicalusefulness," he says.
Hakonarson's team broadened its net, going beyondcherry-picked susceptibility genes to searching a broader collection ofmarkers, including many that have not yet been confirmed, but which reach astatistical threshold for gene interactions or association with a disease.Although this approach embraces some false positives, its overall statisticalpower produces robust predictive results.
By applying a "machine-learning" algorithm that findsinteractions among data points, say the authors, they were able to identify alarge ensemble of genes that interact together. After applying their algorithmto a GWAS dataset for type 1 diabetes, they generated a model and thenvalidated that model in two independent datasets. The model was highly accuratein separating type 1 diabetes cases from control subjects, achieving AUC scoresin the mid-80s.
Hakonarson points out that it is crucial to choose a targetdisease carefully.
"Type 1 diabetes is known to be highly heritable, with manyrisk-conferring genes concentrated in one region—the major histocompatibilitycomplex," he notes.
For other complex diseases, such as psychiatric disorders,which do not have major-effect genes in concentrated locations, this approachmight not be as effective.
Furthermore, the researchers' risk assessment model mightnot be applicable to mass population-level screening, but rather could be mostuseful in evaluating siblings of affected patients, who already are known tohave a higher risk for the specific disease.
Hakonarson says the team's approach is more effective, andcosts less, than human leukocyte antigen (HLA) testing, currently used toassess type 1 diabetes risk in clinical settings.
"We would like to see this test reach the market so we caninform subjects at high risk in a better way and give them more options," notesHarkonarson. "We will measure the impact we will have on clinical care in thefuture."
The researchers used data provided by the Wellcome TrustCase Control Consortium and the Genetics of Kidneys in Diabetes study.Hakonarson's co-authors from The Children's Hospital of Philadelphia were KaiWang, Struan Grant, Haitao Zhang, Jonathan Bradfield, Cecilia Kim, EdwardFrackleton, Cuiping Hou, Joseph T. Glessner and Rosetta Chiavacci, all of theCenter for Applied Genomics; Dr. Charles Stanley of the Division ofEndocrinology; and Dr. Dimitri Monos of the Department of Pathology andLaboratory Medicine. Other co-authors were Constantin Polychronakos and Hui QiQu of McGill University in Montreal; and Zhi Wei of the New Jersey Institute ofTechnology.
