by Simone Salemme
For our June paper of the month, we have chosen EADB., EADI., Bonn. et al. Consensus meta-analysis of genome-wide association studies for Alzheimer’s disease and related dementias. Nat Genet 58, 1214–1225 (2026). https://doi.org/10.1038/s41588-026-02583-1.
Large-scale genetic studies have transformed our understanding of Alzheimer’s disease (AD), but they have also introduced an important methodological challenge. As genome-wide association studies become larger, they increasingly combine clinically diagnosed cases with biobank-defined or proxy cases of Alzheimer’s disease and related dementias (ADRD). This improves statistical power, but it can also make it harder to distinguish genetic signals that are more specific to clinically diagnosed AD from those linked to broader dementia phenotypes.
This month’s selected paper, published in Nature Genetics, addresses this challenge through a major consensus meta-analysis of genome-wide association studies for AD and ADRD. The study brought together European-ancestry data from several large consortia and cohorts. Overall, the analysis included 128,681 cases or proxy cases of ADRD and 849,833 controls or proxy controls.
The study identified 91 genome-wide significant ‘tier 1’ loci associated with ADRD risk. Sixteen of these loci were new in European-ancestry samples at the time of analysis. The authors also identified 25 independent secondary signals across 16 loci, adding further detail to known genetic regions. In addition, they reported 18 ‘tier 2’ loci, 15 of which were new, but classified these as requiring further external validation.
A particularly valuable aspect of the study is its attention to phenotype definition. The authors performed sensitivity analyses excluding proxy ADRD samples and excluding large biobank samples. Among the 91 tier 1 loci, 75 remained genome-wide significant in the no-proxy analysis, while 56 were genome-wide significant in the no-biobank analysis focused on clinically diagnosed AD cases. This distinction is important because it helps separate loci that appear robust across broader ADRD definitions from those most clearly associated with clinically diagnosed AD.
The study also examined biological interpretation. Consistent with previous GWAS findings, genes enriched for ADRD or AD association signals were overexpressed in microglia, and association signals were enriched in pathways related to tau, amyloid, lipids, immunity and endosome/lysosome biology. These findings reinforce the view that AD genetic risk reflects multiple biological processes rather than a single pathogenic pathway.
Another important contribution is the analysis of neuropathology endophenotypes. The authors built polygenic scores based on tier 1 signals, excluding APOE, and tested their associations with 11 neuropathological endophenotypes in the ACT cohort and the ADC/NACC dataset. The main score was primarily associated with AD-related neuropathology rather than non-AD pathology. In the ADC/NACC dataset, individuals in the highest decile of the score had approximately a twofold higher risk of Braak neurofibrillary tangle stage greater than 4 and moderate-to-severe neuritic amyloid plaque pathology at death compared with individuals in the median score group.
However, the paper is careful not to overstate the immediate predictive value of these findings. Although the polygenic score improved discrimination for Braak stage and CERAD score beyond age at death, sex, and APOE ε4 or ε2 alleles, the variance explained remained low. The authors also note that larger datasets with well-characterised AD cases and neuropathological information will be needed to clarify the impact of individual loci on AD pathology versus other neuropathologies.
This paper provides a clearer framework for interpreting genetic findings in an era of very large GWASs, where statistical power, phenotype precision and biological specificity must be considered together. By distinguishing between ADRD, proxy-based phenotypes, biobank-derived samples and clinically diagnosed AD, the study offers a more nuanced map of genetic risk.
