Exploiting biological priors and sequence variants enhances QTL discovery and genomic prediction of complex traits

Table 2 Description of BayesRC models used to analyse the SEQ ^a genotype data

Name of BayesRC Model	Variant Allocation to Classes I, II and III	Number of variants per class^c
BayesRC Seq	I. NSC (non-synonymous coding) II. REG (potentially regulatory) III. CHIP (HD SNP chip variants)	45,026 578,734 370,259
BayesRC Lact	I. NSC & in Lact ^b genes II. All variants other than NSC that overlap Lact gene regions (±50Kb) III. All other SEQ variants not in class I or II	4650 64,518 924,851
BayesRC RLact	I. NSC & in random set of 790 genes II. Variants other than NSC that overlap a random set of 790 genes (±50Kb) III. All other variants not in class I or II	4350 61,748 927,921

^aSEQ = pruned set of 994,019 genome-wide sequence variants from coding and regulatory regions as well as SNP from a high density genotyping array. Variants were allocated to one of three BayesRC classes as listed
^bLact refers to a set of 790 candidate genes shown in an independent study to be differentially expressed in association with altered milk production
^cNumbers generally reduced slightly from those listed because variants with MAF < 0.002 in any given training population were also excluded from the analyses

ISSN: 1471-2164