Background: Current GWAS discoveries have discovered novel clinical improvements in recent decades, such as estimating whole-genome risk. Genetic prediction of traits has substantial impacts on public health care and disease prevention. This study aimed to investigate the effects of different linkage disequilibrium (LD) patterns on genomic prediction accuracy and SNP-based heritability estimation for four lipid profile traits.
Results: This family-based study included 11,798 individuals ranging from 3 to 80 ys, extracted from Tehran Cardiometabolic Genetic Study (TCGS). LD patterns were considered on different thresholds (0.01, 0.03, 0.05, 0.07, 0.09, 0.1, 0.2, 0.3, 0.5, 0.6, 0.7, 0.8, and 0.9) to create subsets of SNPs. We have compared the prediction accuracy and SNP-based heritability estimation of the selected SNPs within these patterns as well as randomly selected SNPs with equal sizes. Subsets of SNPs selected based on LD patterns had a higher prediction accuracy level than subsets of SNPs selected randomly, and when the LD threshold increases, the difference tends to zero. The results were consistent when the prediction accuracy of subsets were adjusted for their SNP numbers in all traits. For all traits, when the number of SNPs was adjusted, between LD threshold 0.01 and 0.2, both prediction accuracy and SNP-based heritability have a dramatic rise. After substantial growth, there was a steady decline, and they reach a peak at an LD threshold between 0.2 and 0.3.
Conclusions: This research indicated that having selected subsets of SNPs based on the LD threshold always outperform randomly selected SNPs for prediction objectives. However, determining the specific LD threshold for prediction purposes might be controversial since achieving the highest level of prediction accuracy, when the number of SNPs is adjusted, prompts different results (in our case, 0.3 when the SNP number was adjusted and 0.9 when the SNP number is not adjusted). Finally, we concluded that choosing the LD threshold as a tool to boost genetic prediction accuracy should be used with intense care.