|Session Title: Complex Traits: Theory and Methods Session Type: Poster|
|Session Location: Exhibit Hall, Level 2, Convention Center Session Time: Wed 10:00AM-7:00PM|
|Program Number: 408W Presentation Time: Wed, Oct 12, 2011, 3:00PM-4:00PM|
|Keywords: Complex Traits: Theory and Methods, KW029 - CLINICAL HISTORY, KW154 - RISK ASSESSMENT, KW143 - PREDICTIVE TESTING, KW076 - GENETIC TESTING, KW105 - MATHEMATICAL MODELING|
Prediction of complex multifactorial disease: comparing family history and genetics. C. B. Do1, J. M. Macpherson1, D. A. Hinds1, B. Naughton1, U. Francke1,2, N. Eriksson1 1) 23andMe, Inc, Mountain View, CA; 2) Stanford University School of Medicine, Stanford, CA.|
In clinical practice, family history is generally considered an important risk factor for a wide variety of complex diseases. Over the last several years, however, commercial SNP-based tests for low penetrance common variants associated with complex diseases have become increasingly available. Although the use of family history and genetic risk prediction is well understood for simple Mendelian disorders, to date, little is known regarding the relative performance of these methods for diseases with complex multifactorial inheritance.
Using quantitative genetic theory, we analyze the influence of disease architecture on the predictive capacity of family history and genetic models with the goal of understanding the range of scenarios in which one of the approaches may be advantageous compared to the other. We find that for extremely common multifactorial disorders (10% prevalence, e.g., type 2 diabetes in the United States), an idealized family history-based algorithm can achieve predictive accuracies on par with SNP-based models that account for 20-30% of the heritability of the disease, meeting or exceeding what SNP-based models have achieved thus far. However, many diseases of important public health consequence are neither extremely common nor extremely rare, occurring in one out of every 100 to 1,000 individuals (e.g., Crohn's disease). For these diseases of moderate frequency, the point of equivalency occurs at only 1-4% of the variance explained, well within the detection limits of GWAS today. These findings suggest that for less common diseases, currently known genetic factors may in fact already be better discriminators of risk than their family history-based counterparts, despite the large fraction of missing heritability that remains to be explained.
Our results provide insight into the influence of disease architecture on risk prediction performance and its implications for clinical practice. In particular, our findings highlight the potential pitfalls of drawing conclusions regarding the relative merits of family history and genetics without considering the broad spectrum of human disorders, and illustrate the importance of considering both family history and genetics in combination when evaluating an individual's predisposition to disease.
You may contact the first author (during and after the meeting) at