Background: Accurate forecasting of Alzheimer's Disease (AD) progression is critical for personalized patient management and clinical trial stratification. However, current predictive models often struggle to effectively integrate high-dimensional neuroimaging with longitudinal clinical data. We introduce AD-LLaVA-3D, a novel multimodal framework designed to bridge this gap by adapting large vision-language models for volumetric and temporal forecasting. Methods: We leveraged the LLaVA-NeXT-Video architecture to treat 3D MRI volumes as temporal sequences, enabling the model to process volumetric imaging alongside longitudinal Tabular Clinical Records (TCR). The model was trained on the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort (n=764) and evaluated using a rigorous patient-level split. We assessed its ability to forecast a suite of future clinical indicators (e.g., CDR-SB, MMSE) against traditional machine learning baselines (Lasso, Random Forest, Gradient Boosting) and specialized deep learning models (ResNet-3D, Med-Flamingo). Results: AD-LLaVA-3D demonstrated superior predictive accuracy on the ADNI test set, achieving a Coefficient of Determination ($R^2$) of 0.68 for the critical CDR-SB score, surpassing the best-performing baseline (R^2=0.66). Crucially, in an independent external validation on the Open Access Series of Imaging Studies (OASIS) cohort (n=76), our model exhibited exceptional generalization (R^2=0.82$, $MSE=0.54), whereas comparison models showed significant performance degradation (R^2 < 0.60). Conclusions: This study presents the first application of a video-based multimodal architecture for AD progression forecasting. By effectively integrating 3D MRI with tabular clinical records, AD-LLaVA-3D offers a robust, generalizable tool for monitoring disease trajectories, significantly advancing predictive capabilities beyond current unimodal or static methods. Highlights: First-in-Class Architecture: We introduce the first application of video-based Large Vision-Language Models (LVLMs) to interpret 3D volumetric MRI as a temporal sequence, capturing longitudinal neurodegeneration more effectively than static 3D-CNNs. Robust External Validation: The model achieved superior predictive accuracy (R^2=0.82) on an independent external cohort (OASIS), demonstrating exceptional generalization beyond the training population (ADNI). Data-Efficient Multimodal Integration: We developed a novel prompting strategy that integrates sparse Tabular Clinical Records (TCR) without artificial imputation, allowing the model to leverage incomplete real-world medical history. Clinical Trial Enrichment: By accurately forecasting future cognitive scores (CDR-SB, MMSE), AD-LLaVA-3D serves as a precise screening tool to identify "rapid progressors" for clinical trials, potentially reducing failure rates in drug development. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement NIH ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: ADNI DATA: https://adni.loni.usc.edu/data-samples/adni-data/ OASIS DATA: https://sites.wustl.edu/oasisbrains/ I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All data produced in the present study are available upon reasonable request to the authors

Our prior research demonstrated that the neuroprotective effects of clonidine can be reversed by yohimbine, an α2-adrenergic receptor antagonist. Additionally, clonidine has been shown to alleviate anxiety-like behavior in rats after bilateral common carotid artery occlusion by reducing the expression of proteins associated with hyperpolarization-activated cyclic nucleotide-gated (HCN) cation channels. However, the specific mechanisms by which clonidine inhibits HCN channels remain unclear. This study was designed to explore the protective effects of clonidine on neurons subjected to oxygen-glucose deprivation (OGD) injury and elucidate the underlying molecular mechanisms via HCN channels. The protective effects of clonidine on OGD-exposed neurons were confirmed by assessing the neuronal viability using a cell counting kit-8 (CCK-8) assay and by measuring the lactate dehydrogenase (LDH) release. Moreover, we identified the signaling pathways most relevant to clonidine’s action. PCR was performed to assess the PKA, AKT, HCN1, and HCN2 gene expressions, and a western blot assay was used to evaluate the related protein expressions of the AC–cAMP–PKA cascade, the PI3K/Akt pathway, and the HCN channels. Clonidine and ZD7288 individually enhanced neuronal viability under OGD, demonstrating neuroprotective effects, with their combination yielding greater benefit. Clonidine upregulated α2A-AR and Nischarin protein levels. Its protection was attenuated by yohimbine (an α2-AR antagonist) and, to a lesser extent, by efaroxan (an I1R antagonist). KT5720, an AC–cAMP–PKA pathway inhibitor, synergized with clonidine and suppressed OGD-induced increases in HCN1 and HCN2 expression. Conversely, LY294002, a PI3K/Akt inhibitor, counteracted clonidine’s protection and further enhanced HCN1/HCN2 expression. These findings indicate that clonidine protects against OGD-induced injury mainly via α2-AR and partially through I1R, potentially by modulating HCN channels via the AC–cAMP–PKA and PI3K/Akt pathways.

Abstract BACKGROUND Understanding of early Alzheimer's disease (AD) progression is critical for timely diagnosis and treatment evaluation, but traditional diagnostic groups often lack sensitivity to subtle early‐stage changes. METHODS We developed a Self‐supervised Longitudinal Progression Embedding (SLOPE) method, an unsupervised dimensionality reduction method that models amyloid progression in AD on a continuous scale that preserves the temporal sequence of follow‐up visits. Applied to longitudinal amyloid positron emission tomography data, SLOPE generates a two‐dimensional trajectory capturing global amyloid accumulation across the AD continuum. RESULTS SLOPE‐derived pseudotime scores better preserved temporal consistency across diagnostic groups and longitudinal follow‐up visits and can be generalized to subjects. The learned trajectory revealed biologically consistent amyloid spreading patterns and greater sensitivity to early progression than global amyloid standardized uptake value ratio. DISCUSSION SLOPE provides a continuous staging of amyloid pathology that complements global amyloid measures by capturing early localized progression.

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by progressive cognitive decline, driven by the accumulation of amyloid-beta plaques, tau tangles, and neuronal atrophy. This study analyzes three imaging modalities corresponding to the three hallmark biomarkers of AD: amyloid PET, tau PET, and structural MRI. Using cortical measurements from the ADNI dataset, we apply PHATE (Potential of Heat-diffusion for Affinity-based Trajectory Embedding), a dimensionality reduction technique, to uncover continuous trajectories of disease progression. We derive pseudotime values from PHATE embeddings via Slingshot, a principal-curve-based pseudotime inference algorithm. In parallel, we apply SuStaIn (Subtype and Stage Inference), a machine learning model that uncovers distinct biomarker event sequences and assigns subjects to discrete disease stages. We observe strong correspondence between SuStaIn-predicted stages and PHATE-derived pseudotimes, indicating temporal coherence across modeling frameworks. SuStaIn further reveals non-overlapping, modality-specific sequences of biomarker abnormalities, consistent with prior neuropathological models. We validate these sequences using pseudotime-aligned kernel density models and clinical measures. Together, our results support the integrative use of these approaches for evaluating AD progression. Future steps will focus on anchoring pseudotime to real-world clinical timelines and experimentally validating SuStaIn-predicted biomarker cascades.

Although cerebrospinal fluid (CSF) biomarkers detect Alzheimer's disease (AD) changes early, their invasiveness and cost limit widespread use. Blood-based metabolite measures, reflecting real-time physiological states, may offer a non-invasive alternative for early AD detection and monitoring. Using data from the Alzheimer's Disease Neuroimaging Initiative, we measured 249 plasma metabolites via NMR spectroscopy and identified those significantly altered in AD compared to controls. We then applied Discriminative Event-Based Modeling (DEBM) to estimate the temporal sequence of AD-related blood metabolites alongside CSF biomarkers (Aβ42, total tau, phosphorylated tau). To validate these findings, we conducted longitudinal analyses assessing cross-correlation with cognitive performance, mixed-effects models relating baseline metabolites to changes in cognition over 72 months, and longitudinal mediation analyses to examine causal pathways. Eight plasma metabolites were significantly altered in AD (Tab. 1). DEBM indicated that two metabolite ratios—phospholipids to total lipids in medium HDL and in large LDL—potentially preceded or coincided with CSF Aβ42 (Figure 1). Cross-correlation analyses showed both baseline CSF Aβ42 and these metabolite ratios were most strongly associated with executive function and memory at 24 months (Figure 2). Linear mixed models revealed that higher phospholipids to total lipids ratio in medium HDL was linked to better cognition and slower cognitive decline. Longitudinal mediation analysis further supported a pathway where the medium HDL ratio preceded CSF Aβ42, which in turn influenced cognitive decline. Our results suggest that specific HDL lipid composition changes may occur very early in the AD cascade, slightly before or concurrent with CSF Aβ42 alterations. These findings support the potential utility of blood metabolite measures, particularly the phospholipids to total lipids ratio in medium HDL, as early AD biomarkers. Further research and replication in independent cohorts are needed to confirm these observations and advance non-invasive AD screening.

BACKGROUND:Although cerebrospinal fluid (CSF) biomarkers detect Alzheimer's disease (AD) changes early, their invasiveness and cost limit widespread use. Blood-based metabolite measures, reflecting real-time physiological states, may offer a non-invasive alternative for early AD detection and monitoring. METHOD:Using data from the Alzheimer's Disease Neuroimaging Initiative, we measured 249 plasma metabolites via NMR spectroscopy and identified those significantly altered in AD compared to controls. We then applied Discriminative Event-Based Modeling (DEBM) to estimate the temporal sequence of AD-related blood metabolites alongside CSF biomarkers (Aβ42, total tau, phosphorylated tau). To validate these findings, we conducted longitudinal analyses assessing cross-correlation with cognitive performance, mixed-effects models relating baseline metabolites to changes in cognition over 72 months, and longitudinal mediation analyses to examine causal pathways. RESULT:Eight plasma metabolites were significantly altered in AD (Tab. 1). DEBM indicated that two metabolite ratios-phospholipids to total lipids in medium HDL and in large LDL-potentially preceded or coincided with CSF Aβ42 (Figure 1). Cross-correlation analyses showed both baseline CSF Aβ42 and these metabolite ratios were most strongly associated with executive function and memory at 24 months (Figure 2). Linear mixed models revealed that higher phospholipids to total lipids ratio in medium HDL was linked to better cognition and slower cognitive decline. Longitudinal mediation analysis further supported a pathway where the medium HDL ratio preceded CSF Aβ42, which in turn influenced cognitive decline. CONCLUSION:Our results suggest that specific HDL lipid composition changes may occur very early in the AD cascade, slightly before or concurrent with CSF Aβ42 alterations. These findings support the potential utility of blood metabolite measures, particularly the phospholipids to total lipids ratio in medium HDL, as early AD biomarkers. Further research and replication in independent cohorts are needed to confirm these observations and advance non-invasive AD screening.

Multi-omics data provides a comprehensive view of biological systems and enables researchers to uncover intricate molecular mechanisms underlying complex diseases. However, multi-omic data is often incomplete and joint modeling of multi-omics data will lead to exclusion of a large portion of subjects. Furthermore, most current multi-omics studies pinpoint individual -omics markers, which may not interact, posing challenges for interpretation. In this study, we developed an interpretable deep trans-omic fusion neural network, TransFuse, to include incomplete -omic data for training of prediction models. When evaluated using the data from two Alzheimer’s disease cohorts, TransFuse generally showed superior or comparable performance over competing methods in a wide range of metrics like classification accuracy and F1 score. In addition, TransFuse yielded a subset of multi-omics features forming functional disease network modules, providing valuable insights into underlying molecular mechanism. In addition, almost all the genetic variants identified by TransFuse are expression quantitative trait locus (eQTLs) specific to frontal cortex tissue, from which the gene and protein expression data were collected. This highlights the great potential of TransFuse in capturing the tissue-specific information flow. Top pathways enriched include VEGF and EPH pathways, both influencing neural development and synaptic formation.

Alzheimer's disease (AD) is a neurodegenerative disorder that results in progressive cognitive decline but without any clinically validated cures so far. Understanding the progression of AD is critical for early detection and risk assessment for AD in aging individuals, thereby enabling initiation of timely intervention and improved chance of success in AD trials. Recent pseudotime approach turns cross-sectional data into "faux" longitudinal data to understand how a complex process evolves over time. This is critical for Alzheimer, which unfolds over the course of decades, but the collected data offers only a snapshot. In this study, we tested several state-of-the-art pseudotime approaches to model the full spectrum of AD progression. Subsequently, we evaluated and compared the pseudotime progression score derived from individual imaging modalities and multi-modalities in the ADNI cohort. Our results showed that most existing pseudotime analysis tools do not generalize well to the imaging data, with either flipped progression score or poor separation of diagnosis groups. This is likely due to the underlying assumptions that only stand for single cell data. From the only tool with promising results, it was observed that all pseudotime, derived from either single imaging modalities or multi-modalities, captures the progressiveness of diagnosis groups. Pseudotime from multi-modality, but not the single modalities, confirmed the hypothetical temporal order of imaging phenotypes. In addition, we found that multi-modal pseudotime is mostly driven by amyloid and tau imaging, suggesting their continuous changes along the full spectrum of AD progression.

Alzheimer's disease (AD) is a progressive neurodegenerative disorder marked by amyloid-beta plaques (A), neurofibrillary tangles (T), and neuronal loss (N), commonly abbreviated A/T/N. Understanding the spatial progression of neurodegeneration is key to predicting disease trajectories and outcomes. Using Positron Emission Tomography (PET) and Magnetic Resonance Imaging (MRI) data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), this study explores the staging and pseudotime of AD patients using A/T/N biomarkers. We collect cortical measurements from ADNI corresponding to A/T/N modalities (Table 1). For each of the three modalities, we apply PHATE, a dimensionality reduction technique for visualizing trajectories of disease progression. We translate PHATE embeddings into 1D pseudotime values using Slingshot, an algorithm that fits principal curves to the embeddings and orthogonally projects the points onto the fitted curves. We independently run SuStaIn, a machine learning algorithm that predicts patient stages and biomarker sequences underlying disease progression. Finally, we compare SuStaIn's stage predictions with pseudotime values from PHATE and Slingshot to test disease progression robustness and generate hypotheses for biomarker events driving underlying disease progression for each imaging modality. We observe that across all three modalities, the predicted SuStaIn stages are closely associated with the pseudotime values inferred by PHATE and Slingshot (Figure 1), with later stages corresponding to higher pseudotime values. Among the three modalities, amyloid PET and tau PET exhibit the strongest trajectory alignments, while MRI-based volume shows a slightly weaker alignment. We also observe agreement between predicted biomarker event sequences from SuStaIn and PHATE pseudotime-informed event-based models. Pseudotimes of amyloid and tau modalities are moderately correlated (r=0.55); however, SuStaIn suggests different biomarker events driving their progressions. Integrative analysis of multiple computational methods allows for higher confidence in disease progression timings. Our results suggest that amyloid and tau-associated biomarkers follow distinct trajectories in the cortex, and that different computational approaches independently arrive at similar results. While this work generates sequences of staging events and pseudotime values of disease trajectories, future work is needed to validate the putative biomarker events and connect pseudotime to real-time disease progression.

Background: There are various molecular hypotheses regarding Alzheimer’s disease (AD) like amyloid deposition, tau propagation, neuroinflammation, and synaptic dysfunction. However, detailed molecular mechanism underlying AD remains elusive. In addition, genetic contribution of these molecular hypothesis is not yet established despite the high heritability of AD. Objective: The study aims to enable the discovery of functionally connected multi-omic features through novel integration of multi-omic data and prior functional interactions. Methods: We propose a new deep learning model MoFNet with improved interpretability to investigate the AD molecular mechanism and its upstream genetic contributors. MoFNet integrates multi-omic data with prior functional interactions between SNPs, genes, and proteins, and for the first time models the dynamic information flow from DNA to RNA and proteins. Results: When evaluated using the ROS/MAP cohort, MoFNet outperformed other competing methods in prediction performance. It identified SNPs, genes, and proteins with significantly more prior functional interactions, resulting in three multi-omic subnetworks. SNP-gene pairs identified by MoFNet were mostly eQTLs specific to frontal cortex tissue where gene/protein data was collected. These molecular subnetworks are enriched in innate immune system, clearance of misfolded proteins, and neurotransmitter release respectively. We validated most findings in an independent dataset. One multi-omic subnetwork consists exclusively of core members of SNARE complex, a key mediator of synaptic vesicle fusion and neurotransmitter transportation. Conclusions: Our results suggest that MoFNet is effective in improving classification accuracy and in identifying multi-omic markers for AD with improved interpretability. Multi-omic subnetworks identified by MoFNet provided insights of AD molecular mechanism with improved details.

Alzheimer's disease (AD) progresses along a continuum and begins many years before symptom onset. Amyloid beta and neurofibrillary tangles are two AD hallmarks that precede changes in cognitive performance and have been utilized in the current framework for classification and staging of AD patients. This framework, however, relies on the dichotomous classification of individual biomarkers (e.g., amyloid positive or negative) and therefore cannot capture the full spectrum of AD progression. In this study, we propose a multi-factorial pseudotime approach that integrates heterogeneous genotype and amyloid imaging data for continuous staging of AD. Our new staging score demonstrated strong capability in differentiating early-stage subjects with distinct progression rates, suggesting its potential in supplementing existing diagnostic framework for early risk stratification.

Alzheimer's disease (AD) is a progressive and irreversible brain disorder that unfolds over the course of 30 years. Therefore, it is critical to capture the disease progression in an early stage such that intervention can be applied before the onset of symptoms. Machine learning (ML) models have been shown effective in predicting the onset of AD. Yet for subjects with follow-up visits, existing techniques for AD classification only aim for accurate group assignment, where the monotonically increasing risk across follow-up visits is usually ignored. Resulted fluctuating risk scores across visits violate the irreversibility of AD, hampering the trustworthiness of models and also providing little value to understanding the disease progression. To address this issue, we propose a novel regularization approach to predict AD longitudinally. Our technique aims to maintain the expected monotonicity of increasing disease risk during progression while preserving expressiveness. Specifically, we introduce a monotonicity constraint that encourages the model to predict disease risk in a consistent and ordered manner across follow-up visits. We evaluate our method using the longitudinal structural MRI and amyloid-PET imaging data from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Our model outperforms existing techniques in capturing the progressiveness of disease risk, and at the same time preserves prediction accuracy.

Alzheimer’s Disease (AD) is a neurodegenerative disorder characterized by progressive cognitive degeneration and motor impairment, affecting millions worldwide. Mapping the progression of AD is crucial for early detection of loss of brain function, timely intervention, and development of effective treatments. However, accurate measurements of disease progression are still challenging at present. This study presents a novel approach to understanding the heterogeneous pathways of AD through longitudinal biomarker data from medical imaging and other modalities. We propose an analytical pipeline adopting two popular machine learning methods from the single-cell transcriptomics domain, PHATE and Slingshot, to project multimodal biomarker trajectories to a low-dimensional space. These embeddings serve as our pseudotime estimates. We applied this pipeline to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset to align longitudinal data across individuals at various disease stages. Our approach mirrors the technique used to cluster single-cell data into cell types based on developmental timelines. Our pseudotime estimates revealed distinct patterns of disease evolution and biomarker changes over time, providing a deeper understanding of the temporal dynamics of AD. The results show the potential of the approach in the clinical domain of neurodegenerative diseases, enabling more precise disease modeling and early diagnosis.

INTRODUCTION:Alzheimer's disease (AD) initiates years prior to symptoms, underscoring the importance of early detection. While amyloid accumulation starts early, individuals with substantial amyloid burden may remain cognitively normal, implying that amyloid alone is not sufficient for early risk assessment. METHODS:Given the genetic susceptibility of AD, a multi-factorial pseudotime approach was proposed to integrate amyloid imaging and genotype data for estimating a risk score. Validation involved association with cognitive decline and survival analysis across risk-stratified groups, focusing on patients with mild cognitive impairment (MCI). RESULTS:Our risk score outperformed amyloid composite standardized uptake value ratio in correlation with cognitive scores. MCI subjects with lower pseudotime risk score showed substantial delayed onset of AD and slower cognitive decline. Moreover, pseudotime risk score demonstrated strong capability in risk stratification within traditionally defined subgroups such as early MCI, apolipoprotein E (APOE) ε4+ MCI, APOE ε4- MCI, and amyloid+ MCI. DISCUSSION:Our risk score holds great potential to improve the precision of early risk assessment. HIGHLIGHTS:Accurate early risk assessment is critical for the success of clinical trials. A new risk score was built from integrating amyloid imaging and genetic data. Our risk score demonstrated improved capability in early risk stratification.

Over the past decade, Alzheimer's disease (AD) has become increasingly severe and gained greater attention. Mild Cognitive Impairment (MCI) serves as an important prodromal stage of AD, highlighting the urgency of early diagnosis for timely treatment and control of the condition. Identifying the subtypes of MCI patients exhibits importance for dissecting the heterogeneity of this complex disorder and facilitating more effective target discovery and therapeutic development. Conventional method uses clinical measurements such as cognitive score and neurophysical assessment to stratify MCI patients into two groups with early MCI (EMCI) and late MCI (LMCI), which shows their progressive stages. However, such clinical method is not designed to de-convolute the heterogeneity of the disorder. This study uses a data-driven approach to divide MCI patients into a novel grouping of two subtypes based on an amyloid dataset of 68 cortical features from positron emission tomography (PET), where each subtype has a homogeneous cortical amyloid burden pattern. Experimental evaluation including visual two-dimensional cluster distribution, Kaplan-Meier plot, genetic association studies, and biomarker distribution analysis demonstrates that the identified subtypes performs better across all metrics than the conventional EMCI and LMCI grouping.

Alzheimer’s disease (AD) is a highly heritable brain dementia, along with substantial failure of cognitive function. Large-scale genome-wide association studies (GWASs) have led to a set of SNPs significantly associated with AD and related traits. GWAS hits usually emerge as clusters where a lead SNP with the highest significance is surrounded by other less significant neighboring SNPs. Although functionality is not guaranteed even with the strongest associations in GWASs, lead SNPs have historically been the focus of the field, with the remaining associations inferred to be redundant. Recent deep genome annotation tools enable the prediction of function from a segment of a DNA sequence with significantly improved precision, which allows in-silico mutagenesis to interrogate the functional effect of SNP alleles. In this project, we explored the impact of top AD GWAS hits around APOE region on chromatin functions and whether it will be altered by the genetic context (i.e., alleles of neighboring SNPs). Our results showed that highly correlated SNPs in the same LD block could have distinct impacts on downstream functions. Although some GWAS lead SNPs showed dominant functional effects regardless of the neighborhood SNP alleles, several other SNPs did exhibit enhanced loss or gain of function under certain genetic contexts, suggesting potential additional information hidden in the LD blocks.

Multi-omic data like genotype, gene expression and protein expression have been increasingly explored in post-GWAS era to interpret GWAS findings and to gain more insight of the Alzheimer’s disease (AD) mechanism. However, each -omics data type is usually examined individually and identified genetic variations, genes and proteins are not necessarily functionally related. We used GWAS genotype, RNA-Seq gene expression, and protein expression data from 133 subjects of ROS/MAP Project for original study and 121 subjects of Mount Sinai Brain Bank (MSBB) for replication (Fig. 1a). The multi-omic data pre-processing steps were shown in Fig. 1b. We proposed a new interpretable deep neural network MoFNet, extended from Varmole [1], to jointly model the prior knowledge of functional interactions using ROS/MAP cohort (Fig. 1c). It aimed to identify a subnetwork of functional interactions from prior knowledge that are both predictive of AD and evidenced by multi-omic data. Particularly, prior functional interaction network was embedded into the architecture of MoFNet in a way that it resembles the information flow from DNA to gene and protein. The proposed model MoFNet significantly outperformed all other state-of-art classifiers when evaluated using multi-omic data from ROS/MAP cohort (Table 1). It yielded three major multi-omic sub-networks related to innate immune system, clearance of mis-folded proteins, and neurotransmitter release respectively (Fig. 2a). One subnetwork includes all the major members of the SNARE complex, an essential mediator of synaptic vesicle fusion. Majority of our identified genes/proteins are related to synaptic activities in neuronal, astrocyte and microglial cells. Differential expression analysis show MoFNet returned more genes and proteins due to functional connections (Fig. 2b). Around 50% of these findings were replicated in another independent cohort (MSBB). Our identified gene/proteins are highly related to synaptic vesicle function. Altered regulation or expression of these genes/proteins could cause disruption in neuron-neuron or neuron-glia cross talk and further lead to neuronal and synapse loss in AD. Further investigation of these identified genes/proteins could possibly help decipher the mechanisms underlying synaptic dysfunction in AD, and ultimately inform therapeutic strategies to modify AD progression in the early stage. [1] Nguyen et al., Bioinformatics , 2021.

The recent multi-omics analysis explores the integration of multiple biological data types, which can suggest a more comprehensive view of biological processes underlying complex diseases, such as Alzheimer’s disease (AD). Various biological networks have been leveraged as prior knowledge with attempt to discovery of more interpretable multi-omics markers [1]. We applied a new graph neural network to take advantage of this rich prior knowledge together with multi-omics data for identification of system-level AD markers. We used GWAS genotype, RNA-Seq gene expression, and protein expression data from the ROS/MAP Project. In total, 133 subjects with full set of -omics features were included in the study (Table. 1). There are 1,751 functionally connected -omics features in the prior biological network, including 186 peptides, 743 unique genes, and 822 single-nucleotide polymorphisms (SNPs). Multi-omic network of functional connections between SNPs, genes and proteins were created based on the REACTOME database, SNP-gene mapping relationship and the SNP2TFBS database [2]. This network was applied to guide the architecture design of the neural network. The input layer consists of all -omics features. A biological drop-connect layer of Varmole was set as for transparent layer that duplicate of protein and gene nodes in the input layer [3]. Links between input layer and transparent layer were added based on the prior multi-omic network. In addition, we also added self-connection links for all genes and proteins. Our proposed integrated multi-omic DNN largely outperforms other state-of-the-art models (Table. 2). With cut-off at 0.00001, 538 features were identified as discriminative. Three largest connected components were observed with more than 30 nodes (Fig. 1), among which 28 SNPs, 309 genes and 31 proteins form the largest connected component. Top hub nodes are proteins PIK3R1, GRB2, FYN. Shown in Fig. 2 is the list of top REACTOME pathways enriched by genes and proteins in the largest connected component [4]. Integrated multi-omic DNN identified multi-mic subnetworks with great predictive power, providing critical functional connections contributing to AD. [1] Xie, et al., BIB, 2021. [2] Fabregat et al., NAR, 2018. [3] Nguyen, et al., Bioinformatics, 2021. [4] Kanehisa, et al., NAR, 2000.

Alzheimer’s disease is the leading cause of brain dementia, along with which substantial failure of organs and mental issues arise. The abundance of AD related data in the current decade has allowed for much advancement in the field using modern machine learning and deep learning techniques to decrypt pathology. Though diagnostic tools have been modeled over such large expanses of data, the black box problem of decoding the significant biological components contributing to the prediction has been overlooked in the research field.