Computational protocol: Genome-wide association analyses for carcass quality in crossbred beef cattle

Similar protocols

Protocol publication

[…] Heritabilities and genetic correlations among the traits were estimated utilizing analytical gradients [] in the REML VCE 6.0 package [] with the model below excluding the allele substitution effect. To account for population stratification in association analysis, the animals were clustered into groups, using pair-wise population concordance test (PPC) at significance level of 0.05 from PLINK v1.07 []. A univariate animal model to estimate the allele substitution effect at each locus, yijkmnpl = + γ1(ageijkmnpl) + γ2(hijkmnpl) + sexi+∑j=16βjbj+ tk + gm + hyn + ap + αlxl + eijkmnpl was used, where yijkmnpl was the phenotype, the overall mean, γ1 and γ2 the regression coefficients for fixed effects age (ageijkmnpl) and individual’s heterosis (hijkmnpl), respectively, being fit as covariates, sexi the fixed effect of sex (male or female), βj the linear regression coefficients of the jth breed, bj the breed proportion of the jth breed (six major breeds being Angus, Charolais, Simmental, Piedmontese, Limousin, Gelbvieh) in the p animal, tk the fixed effect of the kth trial treatment (57 groups), gm the fixed effect of the mth clusters (75 clusters), hyn the random effect of the nth herd of origin by year group (28 groups), ap the random additive genetic effect of individual p, xl the number of copies of the 2nd allele (0, 1, 2) in the genotype for the lth marker, αl the linear regression coefficient (which is also the allele substitution effect) for the lth marker, s the total number of SNP (38,745) included in the analysis, eijkmnpl the random residual effect. Age, heterosis, sex, trial treatment and cluster were assumed to affect all animals equally. Markers were assumed in linkage disequilibrium with quantitative trait loci (QTL) controlling the trait under investigation. Random effects hy, a, and e were assumed uncorrelated with each other. Covariance matrices of the effects were equal to Iσ2hy,Aσa2, and Iσe2, respectively, where I was an identity matrix, and A the additive numerator relationship matrix among the animals. ASREML [] was used to estimate the allele substitution effect, polygenic variance and residual variance.Type I error rate was controlled by the false discovery rate (FDR) proposed by Benjamini and Hochberg []. The calculation of FDR thresholds was derived from the method proposed by Storey [], who estimated the proportion of the p-values, of true null hypotheses, following a uniform distribution on the interval (0, 1). However the numbers of true null hypotheses in this study were estimated using the histogram of p values. The interval (0, 1) was equally partitioned into 20 bins (e.g. bin1 had p values in (0, 0.05], bin2 (0.05, 0.10]… bin20 (0.95,1]). For each trait, n tests were conducted with observed p values distributed as follows: n1, n2… n19, n20 in bin1, bin2… bin19, bin20, respectively, and n=∑t=12oni; nk was the least number of p values among the bins, then 20nk was the estimated number of true null hypotheses, and the FDR threshold was n−20nk100%n . […]

Pipeline specifications

Software tools PLINK, ASREML
Application GWAS
Organisms Bos taurus