!!update 2019 december!!
AD_sumstats_Jansenetal_2019sept.txt.gz are an updated version of the original summary statistics. In the meta-analysis of the previous version, the UKB dataset was weighted with an incorrect sample size, overestimating it by max 22,246 (5%). This results in a slight underestimate of effect sizes and slightly less significant P-values (in the order of the 9th decimal).
We have reran the meta-analysis and have uploaded the new summary statistics. For new analyses we advise to use the updated version of the summary statistics. However, for analyses that have been performed already, we think it is acceptable to work with the previous results. We reran all our functional follow-up results and observed minimal changes, and no alternative conclusions were drawn.
-----------------------------------------------------------------------------------------------------
!!original 2019 january!!
AD_sumstats_Jansenetal.txt.gz are the results from the Phase 3 analysis as desribed in "Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer's disease risk" by Jansen et al in Nature Genetics. The file contains 13,367,301 variants described by the following columns:
uniqID.a1a2 = uniq ID per variant, format -> CHR:BP_A1_A2
CHR = chromosome
BP = base pair position
A1 = allele 1 (effect allele)
A2 = allele 2 (non-effect allele)
SNP = rsID (if available)
Z = Z-statistic
P = p-value
Nsum = sample size by simply summing n per cohort
Neff = effecting sample size
dir = directions of effect per cohort (order: ADSP, IGAP, UKB, PGC-ALZ)
EAF = allele frequency of allele 1
BETA = effect size
SE = standard error
Additional comments for specific columns:
Column 2 to 11 are outputed by the mvGWAMA code, which misses reporting MAF and effect sizes. We therefore added the MAF BETA and SE.
MAF: Minor allele frequencies are initially based on HRC and 1000G phase 3 (match for 12,178,387 variants). For the remaining 1,188,914 variants the MAF was reported based on own datasets (match for 1,131,053 variants). For the subsequent 57,861 variants (0.4% of total dataset) no MAF is reported, as the original IGAP-sumstats miss information on MAF (these variants are all solely based on IGAP).
BETA: $Z/sqrt((2*$MAF*(1-$MAF))*($Neff+ ($Z)^2))
SE: 1/sqrt((2*$MAF*(1-$MAF))*($Neff+($Z)^2))