Obesity Genetic Predisposition Learned with Whole Genome Prediction and Family History

There is great interest in complex trait predictions using genomic profiles (e.g., for personalized medicine). The aim of this research is to predict the Body Mass Index (BMI; kg/m²), using family history and dense genotyping. Data consist of adjusted BMI (with sex, cohort and age) from Framingham Heart Study (FHS). The analysis was performed with Bayesian regressions, for (1) SNP-based predictive model where BMI was regressed on 28,000 genome wide SNP; and (2) a pedigree-based mixed model accounting for relationships among subjects, traditionally used in animal and plant breeding for genetic evaluations (using pedigree information from FHS). Models were trained (n= 8,528), and predictive ability was evaluated in an independent set (n= 488). Predictive ability was evaluated with the correlation between the adjusted BMI (true) and the model prediction in the testing set. Correlations for SNP-based and pedigree-based models were 0.364 and 0.347, respectively. In the pedigree-based model the proportion of variance of the adjusted response that corresponds to the estimated additive genetic effects was 0.42, (being an estimate of the heritability). In the SNP-based model, the proportion of variance with adjusted response regressed on SNP genotypes was 0.54. In summary, our findings suggest that it is possible to explain an important proportion on the genetic variability using dense genetic markers as well as pedigree information. Both models show to be competitive to measure genetic predisposition for BMI and could be extended to other traits with prediction proposes.