Improving Disease Prediction by Incorporating Family Disease History in Risk Prediction Models with Large-Scale Genetic Data

Abstract

Despite the many successes of genome-wide association studies (GWAS), the known susceptibility variants identified by GWAS have modest effect sizes, leading to notable skepticism about the effectiveness of building a risk prediction model from large-scale genetic data. However, in contrast to genetic variants, the family history of diseases has been largely accepted as an important risk factor in clinical diagnosis and risk prediction. Nevertheless, the complicated structures of the family history of diseases have limited their application in clinical practice. Here, we developed a new method that enables incorporation of the general family history of diseases with a liability threshold model, and propose a new analysis strategy for risk prediction with penalized regression analysis that incorporates both large numbers of genetic variants and clinical risk factors. Application of our model to type 2 diabetes in the Korean population (1846 cases and 1846 controls) demonstrated that single-nucleotide polymorphisms accounted for 32.5% of the variation explained by the predicted risk scores in the test data set, and incorporation of family history led to an additional 6.3% improvement in prediction. Our results illustrate that family medical history provides valuable information on the variation of complex diseases and improves prediction performance.

Keywords: Genetic variability in complex binary traits; Liability threshold model; family history; penalized prediction model; risk prediction in complex disease.

Figure 1

A schematic overview of the analysis. Individuals in the main data set include genotyped SNPs, while the external data set of those individuals includes relative’s relationship and disease status. A 10-fold cross-validation scheme was applied to build and test the performance of the prediction models

Figure 2

Clinical variables between cases and controls. Conditional mean (CM) (A), age (B), sex (C), body mass index (BMI) (D), systolic blood pressure (SBP) (E), and diastolic blood pressure (DBP) (F) are shown in boxplots. Two-sample t-test was performed to obtain P-values (* < 0.05; *** < 0.001). For sex, a χ² test was conducted.

Figure 3

Model comparison with different family history measures. Prediction performance (AUC) is depicted without family history (red), with weighted mean (blue), and with conditional mean (green). AUC, area under the curve; EN, Elastic-Net; SCAD, smoothly clipped absolute deviation; T.Ridge, Truncated Ridge.

Figure 4

Proportion of variation explained by each variable in the final model. For six clinical variables [age, sex, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), and conditional mean (CM)], the individual proportions of variation are shown, whereas variation explained by the 5000 SNPs is shown according to their summed proportion.