Lee EH, et al. Dement Neurocogn Disord. 2020 Sep;19(3):96-107. https://doi.org/10.12779/dnd.2020.19.3.96

Background: Plasma phosphorylated tau (p-tau) 217 test has emerged as a minimally invasive and accessible alternative to positron emission tomography imaging and cerebrospinal fluid analysis for Alzheimer’s disease (AD) diagnostics. However, the diagnostic performance of p-tau217 across diverse cognitive and demographic subgroups remains underexplored. This multicentre cross-sectional study aimed to assess the diagnostic utility of plasma p-tau217 using a double cut-off approach in a large, diverse cohort, focusing on subgroup analyses based on cognitive status, age, sex, body mass index and APOE ε4 carrier status. Methods: Plasma p-tau217 levels were analysed in cognitively unimpaired (CU) and cognitively impaired (CI) individuals. Double cut-offs for p-tau217 levels were selected to classify participants into amyloid-negative, intermediate and amyloid-positive groups. Diagnostic performance metrics including sensitivity, specificity, positive predictive value and negative predictive value were evaluated across subgroups, and tailored cut-off strategies were explored for specific populations. Results: The optimal cut-offs differed between CU and CI groups. In the CI group, diagnostic accuracy was consistently high across all subgroups, meeting confirmatory test standards with sensitivity and specificity ≥90%. In the CU group, the appropriate standards varied by subgroup. Participants aged <65 years required alternative cut-offs to improve sensitivity to 85.0% and maintain specificity at 95.7%. Conclusion: Plasma p-tau217 demonstrated robust diagnostic accuracy across CI subgroups and highlighted the importance of tailored cut-off thresholds for CU populations. These findings support the integration of plasma p-tau217 into clinical workflows for AD diagnostics, emphasising its potential for early detection and risk stratification.

GWAS of Alzheimer’s disease have been predominantly based on European ancestry cohorts with clinically diagnosed patients. Increasing the ancestral diversity of GWAS and focusing on imaging brain biomarkers for Alzheimer’s disease may lead to the identification of new genetic loci. Here, we perform a GWAS on cerebral β-amyloid deposition measured by PET imaging in 3,885 East Asians and a cross-ancestry GWAS meta-analysis with data from 11,816 European participants. Our GWAS analysis replicates known loci (APOE4, CR1, and FERMT2) and identifies a novel locus near SORL1 that is significantly associated with β-amyloid deposition. Single-nucleus expression analysis shows that SORL1 is differentially expressed according to β-amyloid positivity in microglia. Our joint association analysis using the SORL1 lead variant (rs76490923) and the APOE4 allele demonstrates that the risk of β-amyloid deposition is reduced by up to 43.5% in APOE4 non-carriers and up to 55.6% in APOE4 carriers, according to the allelic dosage of the rs76490923 T allele. Our findings suggest that SORL1 may play an important role in the pathogenesis of Alzheimer’s disease, particularly in relation to β-amyloid deposition.

Genome-wide association studies (GWAS) on Alzheimer’s disease (AD) have predominantly focused on identifying common variants in Europeans. Here, we performed whole-genome sequencing (WGS) of 1,559 individuals from a Korean AD cohort to identify various genetic variants and biomarkers associated with AD. Our GWAS analysis identified a previously unreported locus for common variants (APCDD1) associated with AD. Our WGS analysis was extended to explore the less-characterized genetic factors contributing to AD risk. We identified rare noncoding variants located in cis-regulatory elements specific to excitatory neurons associated with cognitive impairment. Moreover, structural variation analysis showed that short tandem repeat expansion was associated with an increased risk of AD, and copy number variant at the HPSE2 locus showed borderline statistical significance. APOE ε4 carriers with high polygenic burden or structural variants exhibited severe cognitive impairment and increased amyloid beta levels, suggesting a cumulative effects model of AD risk.

Kang H, et al. Dement Neurocogn Disord. 2025 Apr;24(2):91-101. https://doi.org/10.12779/dnd.2025.24.2.91

Alzheimer’s disease (AD) has complex genetic factors that make accurate prediction from genomic data challenging. Current methods lack confidence estimates for their predictions. Our objective is to develop and evaluate an uncertainty-aware deep learning framework for AD prediction that can identify which predictions are reliable. We developed Transformer-based, Uncertainty-aware, Ensemble Network (TrUE-Net), a deep learning framework that combines transformer and random forest models to predict AD from whole-genome sequencing data. The key innovation is using Monte Carlo Dropout to estimate prediction confidence, allowing the model to identify cases where it is uncertain. We analyzed 1050 individuals (607 AD, 443 controls) from Alzheimer’s Disease Neuroimaging Initiative cohort. On 525 test samples, accuracy of TrUE-Net without uncertainty threshold was 65.1% with area under the receiver operating characteristic curve 0.664. Using uncertainty thresholds, the model classified 75.4% of samples (n = 396) as ‘uncertain’ and 24.6% (n = 129) as ‘certain’. The uncertain group showed accuracy of 62.6% and F1 of 0.584, while the certain group showed accuracy of 72.9% and F1 of 0.821. Uncertainty quantification through Monte Carlo Dropout provides a framework for assessing prediction reliability in genomic AD classification, potentially allowing more informed interpretation of model outputs.