IntroductionEarly prediction of Alzheimer's disease (AD) is of great significance for disease intervention, personalized diagnosis, and treatment. With the rapid accumulation of multi-omics data, integrating complementary information from different molecular layers has become a key challenge in improving AD prediction. However, conventional methods have limitations in fully capturing modality-specific heterogeneity and latent relationships among samples.MethodsTo address these limitations, we propose SDCL-AD, a structure-aware dual-graph contrastive learning framework for AD prediction. SDCL-AD first constructs dual sample similarity graphs using proteomic and transcriptomic features, respectively, and then employs dual-graph autoencoders to learn low-dimensional representations that preserve modality-specific topological structures. On this basis, a cross-modal contrastive learning mechanism is introduced to align the representations of the same sample under different omics views, thereby enhancing the consistency and discriminability of shared disease representations. Finally, the learned structure-aware topological features are fused with the original multi-omics features to complete AD prediction.ResultsSDCL-AD achieved an AUC of 0.9157 and an AUPR of 0.9381, demonstrating superior predictive performance compared to baseline methods. Additional ablation studies and interpretability analyses further validated SDCL-AD's capability in integrating multi-omics data.DiscussionThese results suggest that SDCL-AD effectively integrates multi-omics information while capturing both modality-specific structures and cross-modal consistencies, providing a promising tool for AD prediction.
Linking molecular mechanisms with large-scale brain dynamics remains a fundamental challenge in computational psychiatry. Receptor-informed whole-brain models have provided mechanistic insights into pharmacological interventions, yet they have largely relied on single-receptor topographies that cannot explain the collective response of multiple receptor systems. Here, we combined pharmacological resting-state fMRI with a whole-brain model incorporating cortical expression of adrenergic and dopaminergic receptors to investigate the effects of atomoxetine, a norepinephrine reuptake inhibitor that globally elevates catecholaminergic signaling. Using receptor similarity, we first identified a catecholaminergic chemoarchitecture comprising frontal, posterior and temporal networks, providing the neurochemical scaffold for interpreting atomoxetine-induced brain dynamics. Empirically, atomoxetine selectively stabilized functional connectivity dynamics within the posterior catecholaminergic network. Whole-brain modeling revealed that although individual receptor parameters did not differ between conditions, their collective organization converged toward empirical receptor binding affinities. Remarkably, the simulated atomoxetine state was reconstructed from the placebo configuration through a negative reweighting along this receptor-affinity axis. In silico forward modeling showed that this affinity-driven redistribution gave rise to a cortical modulation pattern aligned with a frontal chemoarchitectural gradient, which was sufficient to simulate the observed dynamical and connectivity changes. Together, these findings suggest receptor binding affinity as a key organizing principle linking molecular receptor organization to brain dynamics, providing a model-based, biologically interpretable computational mechanism through which global catecholaminergic perturbations reshape whole-brain function.
A proper diagnosis of mild cognitive impairment (MCI) is one of the important challenges in the early detection of dementia. For the classification of MCI patients and healthy controls, some electroencephalogram (EEG)-derived indices were calculated and compared. Some studies have been conducted on the high accuracy classification of MCI patients and healthy controls, but classification may depend on the dataset and recording conditions. In this study, we evaluated EEG-derived indices in the Japanese and the Iranian EEG datasets and examined their classification performance under within-dataset and cross-dataset validation. This classifier had common features but was learned for each dataset. The Japanese dataset consisted of 7 MCI patients and 10 healthy controls. The Iranian dataset consisted of 29 MCI patients and 32 healthy controls. The common 16 EEG data were separated into 5 frequency bands. Power spectrum and SL values by Synchronization Likelihood were calculated as features for classification. These standardized and exponentially transformed indices were also used as features. Furthermore, graph metrics were calculated based on SL values before or after standardization and exponential transformation. Features used for classification were selected from these indices. The high classification accuracies were obtained from the classifier using a 34-dimensional feature subset selected from standardized and exponentially transformed EEG-derived indices for lower alpha band. Accuracies of this classifier were 88.24% in the Japanese dataset and 86.89% in the Iranian dataset. To evaluate classification accuracy while minimizing information leakage and optimistic bias, nested leave-one-out cross-validation was performed. Accuracies were 88.24% in the Japanese dataset and 83.61% in the Iranian dataset using exponentially transformed EEG-derived indices for lower alpha band. To evaluate whether the classifier can generalize to independent datasets obtained under different recording conditions, exploratory bidirectional cross-dataset evaluation was performed. Accuracies were 40.98% (Training data: the Japanese dataset, Test data: the Iranian dataset) and 58.82% (Training data: the Iranian dataset, Test data: the Japanese dataset). These findings suggested that resting-state EEG-derived indices for lower alpha band may contain information useful for classification of MCI patients and healthy controls. However, they suggested that classifier based on these indices may partly reflect dataset-specific characteristics.
