Further, Eleftherohorinou et al. [ 17] reported a case where non-significant GWAS markers attained a better performance than that from use of significant GWAS markers alone when predicting rheumatoid arthritis.
As an illustration, a study by Lyssenko et al. [ 5] found that genetic markers performed best when predicting long-term development of diabetes, whereas clinical laboratory assays (such as insulin and glucose levels) performed best when predicting its short-term development.
Modeling the pattern of LD by extension of our formulae would thus be important when many loci are used, as with dense SNP marker maps, or when predicting additive genetic values in other species, such as some livestock populations where the extent of LD is large compared to human [30], [31].
These results indicate that SPLS, BayesB, and to some extent BL are able to select a subset of markers with large effects when predicting SGNC similar to TYM.
Habier et al. [ 30] reported that, with 1000 markers in LD with 10 QTL, GBLUP fitted 100% of SNPs when predicting GEBV, while only a small subset of markers (1.82% to 5.23%) were fitted in Bayesian methods.
This indicates that BayesB and SPLS are able to correctly select the subset of markers with a large effect from all the SNPs when predicting the trait.
L ong et al. (2010) showed with real and simulated data that nonparametric RBF regression methods can outperform Bayes A when predicting total GVs in the presence of nonadditive effects using SNP markers.
The correlation between marker-based prediction accuracy and linkage is modestly positive (r = 0.13) when using 200 features and, as expected, significantly greater when predicting on the single best feature alone (r = 0.48).
For example, even with as few as 96 markers, where the gains in GEBV accuracy were 0.1 0.2 in the maize population when training on all phenotypes, there was no accuracy gain when predicting unphenotyped lines.
Therefore when predicting the H. sapiens M.
