We conducted replication analyses in N=15,496 European subjects from three AD case-control samples (Treutlein et al., 2009; Edenberg et al., 2010; Frank et al., 2012; Gelernter et al., 2014) and one population sample (Heath et al., 2011). Details of the individual samples, genotyping and imputation are provided in Supplementary Information. We conducted look-up analysis of replication SNPs in each sample and meta-analysis of all replication SNPs first in the four ascertained case-control samples, then adding the unascertained population sample, for which AD diagnoses were derived rather than directly assessed (Heath et al., 2011), using METAL (Willer et al., 2010). We weighted meta-analyses by sample size and direction of effect because MQLS does not generate standard errors. Many markers are non-independent due to LD, so we assessed the 274 replication SNPs for independence using SNAP (Johnson et al., 2008) to determine the number of independent tests.
