esidual effects being sampled in different replicates. Each bootstrap sample was reanalyzed using the same Bayes C model used for the real data, and the genetic variances of the SNP window corresponding to the QTL were accumulated for comparison with the test statistic represented by the genetic variance of 
the SNP window identified in the analysis of the real data. If just one bootstrap statistic from the 1000 simulated exceeded the test statistic from the real data, the comparison-wise P-value was determined to be 0.001 < P < 0.002. Multiple testing was taken into account considering the proportion of false positives. This approach controls the proportion of false positive conclusions across all tests undertaken, rather than the probability of making one mistake over all tests, as would be the interpretation of an experiment-wise error correction. The proportion of false positives is calculated as a function of the average comparison-wise Type I error rate, the proportion of true null hypotheses tested among all hypotheses tested, and the power of the test. Candidate PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19801058 gene search Intervals for candidate gene searches were defined as the 5-SNP window plus the 300 kb flanking regions in QTL peaks. Annotated genes contained in this window were identified using Ensembl BioMart tool and the UCSC Genome Browser with the Sus scrofa Build 10.2 assembly. Nonneman et al. BMC Genetics 17:50 Page 4 of 9 Linkage disequilibrium analysis Linkage disequilibrium was estimated for the SNP using Haploview 4.0 software . Haplotype blocks were based on pairwise LD values. Results and discussion In the present study, a GWAS using the PorcineSNP60 BeadChip was performed by means of Bayes C model averaging with random SNP effects for age at puberty. The variances, heritability, and the proportion of total variance explained by the markers are shown in Method MTDFREML GenSel Bayes C 1 from 9396 Mb, SSC10 from 35.637.7 Mb and on SSC16 from 2627.3 Mb. The region on SSC2 at 98 Mb was previously associated with delayed puberty in pigs. Sixteen of the QTL were Halofuginone manufacturer within 0.5 Mb of previously reported associations from the same population where attainment of puberty was considered a categorical trait and a casecontrol experimental design was implemented. Animals from the previous study that had not reached puberty were not included in the current study as their age at puberty was unknown. Two associations were located in published QTL for age of puberty in Meishan x European pig resource populations on SSC1 at 292 Mb and on SSC10 at 68.5-69.5 Mb. Eight regions corresponded to those previously reported by Tart et al. in a genome-wide association study. Candidate genes in QTL regions The most significant QTL accounting for 9.7 % of the genetic variance was located on SSC12 at 15 Mb within the growth hormone and chorionic somatomammotropin hormone gene cluster. Growth hormone and its receptor are necessary for the onset and normal course for attainment of puberty. There are age related changes in somatotropin secretion in pigs, with serum concentrations of GH declining as gilts approach pubertal age. A second QTL accounting for over 7.1 % of the genetic variance was located on SSC7 at 75 Mb near the PRKD1 and C14orf 23 loci. PRKD1 is associated with body-mass index in humans and an SNP in this window near C14orf23 was previously reported to be associated with delayed puberty in relatives of the same population of pigs. For regions previously reported by Tart et al. likely