This study examined associations between functional genetic variation and resting-state EEG features in individuals along the mild cognitive impairment (MCI) continuum, with the aim of identifying genetically informed endophenotypes relevant to early Alzheimer’s disease (AD) pathophysiology. The analysis combined eyes-closed EEG metrics with genome-wide genotype data and used Bayesian reduced rank regression to relate genes to latent EEG features.
The analysis included 169 Finnish participants with cognitive symptoms ranging from subjective cognitive decline to MCI. From eyes-closed resting-state recordings the authors derived periodic spectral features focusing on alpha- and beta-band activity and aperiodic spectral parameters intended to capture broadband, non-oscillatory components of the EEG signal.
Genome-wide genotypes underwent quality control, imputation and functional annotation. The annotated dataset comprised 16,935 gene regions and 90,607 functional variants. These gene-centric variant sets were the basis for association testing against latent EEG features.
Associations between gene-region variation and latent EEG features were tested using Bayesian reduced rank regression (BRRR). BRRR was used to model relationships between multivariate EEG-derived phenotypes and aggregated functional genetic variation, yielding candidate gene associations for periodic and aperiodic EEG features.
At a nominal threshold of P < 0.005, the authors report 145 genes associated with periodic EEG features and 39 genes associated with aperiodic EEG features. The source specifies that none of these associations survived correction for multiple testing. The reported candidate genes were further examined in the context of known AD loci and relevant neurobiology.
Candidate genes associated with periodic EEG features converged on pathways related to synaptic transmission, neuronal excitability and neurodevelopment. Specific genes noted in the source that map to previously AD-associated loci include RASGEF1C and TREML2. These findings suggest a link between genetic variation affecting neuronal signaling and oscillatory EEG characteristics.
The aperiodic-feature candidate genes were relatively more enriched for neuroinflammatory processes; an example given in the source is C5AR2, which resides within a FinnGen region associated with AD. By contrast, periodic-feature candidates were more often involved in intracellular neuronal maintenance and signaling. The authors interpret these differences as indicating that periodic and aperiodic EEG features capture partly distinct biological processes within the MCI continuum.
The authors performed unsupervised clustering of participants based on variants associated with periodic EEG features. These genetically informed clusters differed significantly in plasma p-tau217 concentrations and delayed verbal recall after correction for multiple testing. The top genes contributing to cluster formation included ACAN, INPP5B, CAMKK2, CABIN1 and SPATA13. The source links these cluster differences to AD-related biomarker variation and memory performance, supporting the use of EEG-derived endophenotypes to parse heterogeneity in early cognitive decline.
The source explicitly states that none of the gene associations remained significant after multiple-testing correction, a major limitation for interpreting individual gene results. Data sharing is restricted by the ethics committee and Finnish regulations: raw brain imaging data cannot be openly shared, although individual models derived from the data and the code used for machine learning analyses are available at the BioMag Laboratory GitLab repository. Ethical approval was provided by the HUS Regional Medical Research Ethics Committee, Helsinki University Hospital, and participant consent procedures were followed as required.
The reported results suggest that EEG-derived endophenotypes—periodic oscillatory measures and aperiodic spectral parameters—may index partly distinct genetic and biological processes relevant to early AD pathophysiology, including synaptic/neurodevelopmental and neuroinflammatory pathways. Although associations did not survive multiple-testing correction, the convergence of candidate genes on biologically plausible processes and the link between genetically defined EEG clusters and plasma p-tau217 and delayed verbal recall support further investigation of EEG features as genetically informed biomarkers in prodromal AD. Future work should aim to replicate these findings in larger cohorts and test robustness after stringent correction for multiple comparisons.