GWAS using ssGBLUP

Abstract Text: This study aimed to compare results of genome-wide associations obtained from various methodologies for GWAS when applied to two lines of broiler chicken. Each line contained >250k birds with 3 traits and ~5k SNP60k genotypes. Methods included single-step GWAS, single marker model and BayesB. Mannhattan plots were based on variances of 20-SNP segments, as shorter segments produced noisy plots. Only a few segments explained >1 % of the additive variance. One segment explained >20% variance in BayesB but 3% with ssGWAS and <1% with a single marker model. In two lines, no major segment overlapped for any trait. When analyses used slices of generations (1-3,2-4,3-5,1-5), variances for the same segment varied greatly.  The plots were more distinct with a new data set that included >16k genotypes, but no segment explained >1% of the variance.  Strength of associations strongly depends on methodologies and details of implementations. 

Keywords:

ssGBLUP
GWAS
SNP variance