Abstract
Genetic Analysis Workshop 18 provided a platform for evaluating genomic prediction power based on single-nucleotide polymorphisms from single-nucleotide polymorphism array data and sequencing data. Also, Genetic Analysis Workshop 18 provided a diverse pedigree structure to be explored in prediction. In this study, we attempted to combine pedigree information with single-nucleotide polymorphism data to predict systolic blood pressure. Our results suggested that the prediction power based on pedigree information only could be unsatisfactory. Using additional information such as single-nucleotide polymorphism genotypes would improve prediction accuracy. In particular, the improvement can be significant when there exist a few single-nucleotide polymorphisms with relatively larger effect sizes. We also compared the prediction performance based on genome-wide association study data (ie, common variants) and sequencing data (ie, common variants plus low-frequency variants). The experimental result showed that inclusion of low frequency variants could not lead to improvement of prediction accuracy.
Highlights
Genomic prediction is an important problem in genetics
E ∼ N(0, σe2I), where y ∈ Rn×1 is the response vector; X ∈ Rn×d is the matrix of covariates, including the intercept and other covariates, such as age and gender; β is the vector for regression coefficients of the covariates; G ∈ Rn×p is the genotype matrix and α is the coefficient vector for p single-nucleotide polymorphisms (SNPs); u is the random effect from N(0,σu2K); and e is the residual error with variance σe2
Penalized linear mixed model There is a difficulty in applying the model when d+p +2>n,ie, the number of parameters exceeds the number of samples (d is the number of covariates, p is the number of SNPs treated as fixed effects, 2 is the number of variance components)
Summary
Genomic prediction is an important problem in genetics. It aims at predicting a phenotype outcome based on information from genetic markers, population, pedigree structures, and other relevant covariates. A larger sample size is needed to estimate those small effects more accurately. A larger sample size leads to the improvement of prediction accuracy. Low frequency variants (minor allele frequency [MAF] ≤5%) have not been directly observed in genome-wide association studies (GWAS). The contribution of these lowfrequency variants has not been taken into account in predictive models, which may result in the loss of prediction accuracy
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
