5% for LS and FN BMD, respectively. Genotyping The discovery sample was genotyped via the Infinium assay (Illumina, San Diego, CA) with Human610-quad chip, including 564,214 SNPs. PLINK (version 1.04) was used for data management and quality-control statistics. ICG-001 in vitro Seven hundred seventy-eight individuals

and 488,853 SNPs were retained for analysis following application of strict quality-control criteria. Proteasome function Subjects were excluded according to the following criteria: (1) genotyping call rate <95% (n = 5), (2) autosomal heterozygosity <27% or >31% (the same five subjects with low genotyping call rate), (3) being related or identical to other individuals in the sample (n = 7), and (4) discordance of observed gender and estimated sex (n = 3). SNPs were excluded if (1) genotyping RG-7388 in vivo call rate was ≤95% (1,158 SNPs), (2) Hardy–Weinberg equilibrium p value <1.0 × 10 −4 (904 SNPs), (3) minor allele frequency <0.01 (73,589 SNPs). The average genotyping call rate of retained SNPs was 99.91%. In silico gene-based GWAS of European subjects All GWA data for spine and hip BMD of dCG were

retrieved from the publication [2]. Phenotype modeling To be comparable and compatible in meta-analysis, both studies used standardized residuals of raw BMDs following adjustment for age, weight, and sex (dCG) in the linear regression model. Definition of gene locus We defined the gene locus by its position based on UCSC and ±50 kb. Fifty kilobytes is the mean distance between the top intergenic SNPs and the nearest genes identified in the latest meta-analysis of GWAS of BMD [1]. SOX6 and MEF2C were excluded from the mean distance calculation as they were the outliers.

Statistical analysis In the genome-wide association study, the association test of SNPs with standardized residuals of BMD was implemented via PLINK (version 1.04). For each SNP, the asymptotic Adenosine triphosphate p value for the relationship between the number of minor alleles and BMD was derived from a two-sided t statistic assuming the minor allele has an additive effect. We identified disease-associated genes with two stages. The first stage was the gene-based test using a simulation approach that took account of LD structure of different populations based on HapMap phase two information. Gene-based analysis in each population was done using VEGAS [4]. In brief, each SNP was assigned to each of 17,787 autosomal genes according to positions on the UCSC Genome Browser hg18 assembly. In order to capture regulatory regions and SNPs in LD, the gene boundaries were defined as 51 kb of 5′ and 3′ UTRs. Then, for a given gene with n SNPs, association p values were first converted to upper tail chi-squared statistics with 1 degree of freedom (df). The observed gene-based test statistic was then the sum of all (or a pre-defined subset) of the chi-squared 1 df statistics within that gene. In the current study, we summed the statistics of all SNPs located within a gene.