Linking COVID-19 to Neurodegeneration: A Single-Cell Deep Learning Study of PBMCs in Multiple Sclerosis and Alzheimer’s Disease

Linking COVID-19 to Neurodegeneration: A Single-Cell Deep Learning Study of PBMCs in Multiple Sclerosis and Alzheimer’s Disease | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Linking COVID-19 to Neurodegeneration: A Single-Cell Deep Learning Study of PBMCs in Multiple Sclerosis and Alzheimer’s Disease Asiyeh Mirzaei Koli, Shokoofeh Ghiam, Mohammad Shirinpoor Kharf, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7342199/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract COVID-19 has increasingly been linked to neurological complications that may overlap with those observed in neurodegenerative and autoimmune diseases. In this study, we analyzed single-cell RNA-sequencing data from peripheral blood mononuclear cells (PBMCs) of patients with COVID-19, multiple sclerosis (MS), and Alzheimer’s disease (AD). Using a deep neural network combining autoencoders and adversarial learning, we uncovered distinct and shared transcriptional signatures across these conditions. Top-ranked genes—including HLA-DRB5 , XIST , and DDX3X —were not necessarily differentially expressed but demonstrated strong functional relevance through pathway enrichment and protein interaction analysis, highlighting latent biomarkers often missed by traditional DEG-based methods. Importantly, these candidate genes may aid in the detection of MS and AD among individuals with severe COVID-19 and a family history of these disorders, offering a non-invasive strategy for risk stratification and early intervention. Our findings underscore the value of PBMC-based scRNA-seq and deep neural network frameworks for discovering non-invasive biomarkers and highlight systemic and neuroinflammatory pathways that may connect COVID-19 to long-term neurological outcomes. This integrative approach may pave the way for novel diagnostic and therapeutic strategies, emphasizing the shared immunological underpinnings of these complex diseases. Health sciences/Biomarkers Biological sciences/Computational biology and bioinformatics Health sciences/Neurology Biological sciences/Neuroscience Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1 Introduction SARS-CoV-2, the virus responsible for coronavirus disease 19 (Covid-19), has caused a global health crisis with widespread and persistent symptoms in affected individuals. Approximately 80% of Covid-19 patients experience long-term complications, including fatigue, anosmia, lung dysfunction, and various neurological disorders 1 . Among these symptoms, the neurological effects are particularly concerning due to their potential long-term impact. Emerging research indicates that Covid-19 not only induces immediate neurological symptoms but may also have lasting effects on the central nervous system (CNS), particularly in individuals with pre-existing neurological conditions or those at risk of developing them 2 . For example, recent studies suggest a potential link between Covid-19 and MS, although the mechanisms are still unclear. Case reports describe new MS diagnosis or symptom onset following SARS-CoV-2 infection, such as a 47-year-old woman who developed symptoms three weeks post-infection 3 . Additionally, 51% of MS patients experienced symptom worsening, and 20% developed new symptoms during or after infection. Relapse rates also increased among relapsing-remitting MS patients during the Covid-19 "at-risk" period 4 . Immune activation and inflammation within the CNS may be primary contributors to the neurological symptoms observed in SARS-CoV-2 patients. There are at least three possible routes through which SARS-CoV-2 can invade the brain: Infection in the nasal cavity could allow the virus to directly spread to the olfactory bulb via the olfactory nerves 5 . The virus may infect the eyes and subsequently reach the occipital cortex through the optic nerve 6 . High viral titers in the respiratory tract could enable the virus to enter the bloodstream, spreading to multiple organs, including the CNS 7 , 8 . Replication of the SARS-CoV-2 virus within host cells triggers an immune response by activating toll-like receptors ( TLRs ) and MDA5 or RIG-I , leading to the production of interferons (INF) 9 (Fig. S1 -a). In response to interferon release, a group of immune cells accumulates in the peripheral blood, activating the JAK-STAT1/2 signaling pathway and contributing to the formation of a cytokine storm 10 (Fig. S1 -c, d). This immune response can ultimately compromise the blood-brain barrier (BBB), allowing immune cells to infiltrate the CNS. Once in the CNS, these immune cells initiate severe immune responses, contributing to neurodegeneration and apoptosis, which can further stimulate multiple long-lasting immune responses within the CNS 11 (Fig. S1 -e). Before entering the brain, viral pathogens must penetrate the blood-brain barrier (BBB) 12 . SARS-CoV-2 may cross the BBB through several mechanisms, including the ACE2-mediated transcellular pathway, disruption of the BBB's tight junctions via the paracellular pathway, or intracellular transport, where viral particles are carried by infected cells, bypassing the BBB 13 . Once inside the CNS, the virus can trigger increased inflammatory responses, potentially leading to neurodegenerative diseases, particularly Multiple Sclerosis (MS). Multiple sclerosis (MS) is a chronic inflammatory and demyelinating disease of the CNS, marked by immune cell infiltration, myelin destruction, and axonal damage, leading to neurological disability. Epstein-Barr virus infection is a key driver of MS, establishing a causal link between infection and neurodegeneration 14 . MS onset involves peripheral immune activation, followed by CNS invasion, supporting the "outside-in" theory. Mast cells disrupt the blood-brain barrier (BBB), allowing peripheral T cells, aided by myosin genes and MMPs, to infiltrate the CNS. Once inside, dendritic cells and CNS-resident cells release cytokines (TNF-α, IFN-γ, IL-17), which activate microglia and macrophages, causing myelin and axonal damage via neurotoxic molecules (MMPs, TNF-α, ROS, RNS) and CD8 + T cell cytotoxicity. Astrocytes and NK cells, usually regulatory, also contribute to this process. As the BBB further weakens, more immune cells infiltrate, intensifying neural damage. This immune response accelerates MS progression and may explain how Covid-19 could worsen neurodegeneration 15 . Similarly, Alzheimer’s disease (AD) is characterized by cognitive decline and neuronal loss, with genetic variants linked to immune response and neuroinflammation, such as OAS1 , increasing the risk for both AD and severe Covid-19 16 . Covid-19 can trigger cytokine storms, elevating pro-inflammatory cytokines like IL-1β , IL-6 , and TNF-α , which may contribute to synaptic dysfunction and neurodegeneration, particularly increasing the risk of AD in elderly individuals 17 . The interplay between neuroinflammation and immune response, as seen in both AD and Covid-19, highlights the need for a deeper understanding of how these conditions influence brain health. Single-cell RNA sequencing (scRNA-seq) has significantly advanced our understanding of neurodegenerative disorders by revealing gene expression changes at the single-cell level, identifying key molecular alterations during disease progression 18 . Machine learning models help analyze these complex datasets, enabling precise biomarker identification 19 . While traditional approaches rely on invasive procedures involving brain tissue or cerebrospinal fluid (CSF), they may miss subtle gene expression changes. Recently, there has been growing interest in using blood-based biomarkers for diseases like MS and AD. Blood samples offer a non-invasive alternative, promising earlier diagnosis, safer procedures, and improved disease management 20 , 21 . Recent advances in blood-based biomarker discovery for AD provide a relevant framework for studying the neurological implications of COVID-19. In particular, the GeneDX-PBMC model, introduced by Talebi et al 22 , demonstrated how single-cell RNA sequencing (scRNA-seq) data from peripheral blood mononuclear cells (PBMCs) can be combined with a deep learning framework—integrating autoencoders, classifiers, and discriminators—to identify both differentially expressed and subtle, non-differentially expressed genes associated with AD. This approach enabled the detection of systemic immune alterations linked to neuroinflammation, a mechanism that is also implicated in post-COVID neurological sequelae. Drawing on these methodological insights, our COVID-19 analysis adapts the same computational framework to investigate transcriptional signatures in PBMCs, aiming to uncover shared and disease-specific pathways connecting COVID-19 to neurodegenerative processes. In this paper, we leverage scRNA-seq profiling to explore the genetic links between Covid-19 and neurodegenerative diseases, specifically MS and AD. Traditional biomarker discovery methods, which rely on brain tissue, are not only invasive but also time-consuming and expensive. Furthermore, previous studies have primarily focused on differentially expressed genes (DEGs), often overlooking genes with subtle expression differences that may still play crucial roles in disease progression. To address these limitations, we analyzed scRNA-seq data from PBMCs of patients with MS, AD, and Covid-19. By employing a deep learning approach, we captured hidden nonlinear relationships between genes, effectively analyzed high-dimensional genomic data, and accurately ranked significant genes. Our model was applied separately to four cell types—T-cells, B-cells, NK-cells, and monocytes—identifying both DEGs and significant non-DEGs in each cell type. This provided deeper insights into the molecular mechanisms by which Covid-19 may contribute to these two neurodegenerative diseases. 2 Materials and Methods 2.1 Data collection The publicly available scRNA-seq data for severe Covid-19, MS, and AD, with accession numbers GSE166992 23 , GSE138266 24 , and GSE181279 25 , respectively, were downloaded from the Gene Expression Omnibus (GEO) database 26 . Data for these conditions were collected from PBMCs using the same platform, ensuring comparability in the analysis. Additionally, a Covid-19 scRNA-seq dataset containing mild and asymptomatic cases, along with a severe flu dataset (under the GSE149689 accession number) 27 , was also included for comparative analysis. 2.2 Data preprocessing For quality control across all scRNA-seq datasets, genes with at least 200 features and non-zero counts, along with cells with a minimum count of three, were selected. Low-quality cells were removed from the dataset. After filtering, the data were normalized using a log-transformation to stabilize variance across the datasets. This preprocessing was performed using the Seurat package (version 5.1.1) 28 in R (version 3.6.1) 29 . The results of the preprocessing steps are summarized in Table 1 . Table 1 Summary of Data Quality Control and Gene Statistics Across Different Conditions Condition Number of Samples Number of Raw Cells Number of Cells after QC Number of Cells After Doublet Removal Total Number of Genes Covid-19 (Severe) D:5 N:3 63895 37910 31966 33538 Multiple Sclerosis D:5 N:3 58300 49848 41979 22164 Alzheimer’s Disease D:3 N:2 36849 32177 27562 32738 Covid-19 (Mild & Asymptomatic) D:5 N:4 53962 34607 29474 33538 Flu (Severe) D:5 N:4 39245 24697 21328 33538 This table presents the number of samples, raw cells, cells after quality control (QC), cells after doublet removal, and the total number of genes for each condition. The dataset includes samples from severe and mild cases of Covid-19, MS, AD, and severe flu. The conditions are categorized as Diseased (D) or Normal (N). QC and doublet removal steps were applied to ensure data quality before normalization using the Seurat package. 3 Method 3.1 Overview: Our study employs a robust deep learning framework, directly adapted from the GeneDX-PBMC 22 methodology originally developed for Alzheimer’s disease biomarker discovery, to analyze PBMC scRNA-seq data and uncover gene expression patterns linked to neurodegenerative diseases. After data preparation, we clustered cell types into T cells, B cells, NK cells, and monocytes, identifying significant genes within each cluster (Fig. 1 -A). The framework integrates an autoencoder for feature extraction, a discriminator for adversarial training, and a classifier for disease prediction. The autoencoder reduces data dimensionality, the discriminator distinguishes between real and generated latent vectors, and the classifier predicts normal and diseased classes (Fig. 1 -B). While the architecture, training strategy, and gene-ranking approach remain identical to the original AD study, the input data and subsequent analyses are specific to the present COVID-19 investigation. We applied the model to publicly available PBMC scRNA-seq datasets from patients with COVID-19 (severe, mild, and asymptomatic), along with comparative datasets for multiple sclerosis (MS) and Alzheimer’s disease, to identify transcriptional patterns relevant to COVID-19 pathophysiology and its potential neurological impact. 3.2 Data Preparation To rigorously evaluate our deep learning model, we employed two validation sets. Validation 1 involved splitting the dataset into 80% training and 20% testing, ensuring that the test set remained unseen during training. Validation 2 was performed by shuffling the dataset and excluding one diseased and one normal sample, which were used solely for testing model performance. This dual validation strategy allowed for comprehensive evaluation of the model's ability to generalize to unseen data. 3.3 Gene Selection and Cell-Type Clustering Following the setup of the training and validation sets, we shifted our focus to the training sets to identify significant genes critical for understanding disease progression. After removing batch effects using the IntegrateData function of the Seurat package, we employed K-means clustering to group similar cells, testing various values of k to determine the optimal number of clusters. The best fit was achieved with k = 4, effectively capturing clustering patterns observed across different values of k. We consistently observed that similar normal and diseased cells were grouped within the same clusters. To further refine these clusters, we applied the SingleR package 30 , which independently classified cells into four main types: T cells, B cells, NK cells, and monocytes, as depicted in Fig. 1 -A. This classification aligned well with the clustering results from K-means. Within each cell type and disease, gene selection was performed using training set, applying an adjusted p-value threshold of 0.01. This process included both differentially expressed genes (DEGs) and non-differentially expressed genes (non-DEGs). Subsequently, a vector containing the expression levels of the significant genes was allocated to each cell. This vector was input into the neural network model, which was run independently for each cell type and disease, to identify specific patterns of gene expression. 3.4 Algorithm Our deep learning model integrates three distinct modules: an autoencoder, a classifier, and a discriminator, which collectively perform feature extraction, classification, and adversarial training using the PyTorch 31 library in Python In the feature extraction phase, the autoencoder compresses input data into a lower-dimensional latent space, focusing on deep feature extraction and dimensionality reduction, while the decoder reconstructs the original data from this compact representation. The classifier module then takes these latent representations to generate class predictions, incorporating a loss term that penalizes classification errors and enhances accuracy by prioritizing feature extraction over raw data preservation. Finally, the discriminator module is employed for adversarial training, learning to differentiate between real encoded representations and fake randomly generated latent vectors, thus ensuring the encoder produces meaningful features. This integrated approach facilitates precise classification, effective feature extraction, and robust adversarial training. Each component of the model is explained in detail below. 3.5 First Module: Feature Extraction In this module, the encoder and decoder work together to extract features and achieve dimensionality reduction, producing a more compact representation of the original data. Encoder: The encoder compresses the input data into a lower-dimensional latent representation (Fig. 1 -B-a). This feature vector captures essential patterns of the original data while reducing its dimensionality. It consists of fully connected (linear) layers with batch normalization applied after each layer to enhance training stability and speed up convergence. Rectified Linear Unit (ReLU) 32 activation functions introduce non-linearity, and dropout is used as a regularization technique to prevent overfitting. Decoder: The decoder reconstructs the input data from the latent representation (Fig. 1 -B-b). The objective during training is to minimize reconstruction loss, the difference between the original and reconstructed data. By reducing this loss, the autoencoder learns to focus on the most informative features. The decoder mirrors the encoder's structure, using fully connected layers with batch normalization and ReLU activation functions. Dropout is also applied to prevent overfitting. A final Sigmoid activation function ensures output values fall within the [0, 1] range. The Mean Absolute Error (MAE) loss module measures the absolute differences between the original and reconstructed data, guiding the training process by penalizing the model based on reconstruction errors. MAE = \(\:\frac{1}{n}{\sum\:}_{i=1}^{n}\left(yᵢ-ŷᵢ\right)\) (1) 3.6 Second Module: Adversarial Training Classifier The classifier processes the encoded representation to generate class predictions (Fig. 1 -B-b). It consists of fully connected layers, with batch normalization applied to stabilize training and enhance convergence. Each fully connected layer is followed by a ReLU activation function to introduce non-linearity and capture complex patterns. To mitigate overfitting, a dropout layer is used after the initial layer. The final layer employs a LogSoftmax activation function to transform raw scores into a probability distribution suitable for multiclass classification tasks. The classifier uses cross-entropy loss to measure the difference between the predicted probability distribution and the true class distribution, guiding the model to minimize this difference during training. Cross Entropy Loss = - \(\:{\sum\:}_{i=1}^{n}yᵢlog\left(ŷᵢ\right)\) (2) 3.7 Third Module: Discriminator The discriminator is utilized for adversarial training, a technique often employed in generative adversarial networks (GANs) and applied here in ADEP 33 , 34 . Its primary objective is to differentiate between encoded representations produced by the encoder (real representations) and randomly generated latent vectors (fake representations) (Fig. 1 -B-b). During training, the discriminator learns to distinguish between these real and fake representations. The discriminator's loss is typically calculated using binary cross-entropy or logistic loss, which measures the discrepancy between the discriminator’s predictions and the actual labels (real or fake). This loss functions as a quality control mechanism, ensuring that the encoder produces meaningful and informative representations of the input data. The discriminator consists of multiple fully connected (linear) layers, with batch normalization and ReLU activation functions applied after each layer to enhance training stability and capture complex patterns. The final layer uses a Sigmoid activation function to output a probability score within the range [0, 1], reflecting the likelihood that the input represents real encoded data. The Binary Cross-Entropy (BCE) loss module is employed to quantify the difference between the discriminator's predictions and the true labels. This loss guides the adversarial training process, helping the discriminator more effectively distinguish between real and generated representations. The BCE loss, also known as discriminator loss, is computed using the following formula: BCE = - \(\:\frac{1}{n}\) \(\:{\sum\:}_{i=1}^{n}\left[yᵢ\cdot\:log\left(ŷᵢ\right)+\left(1-yᵢ\right)\cdot\:log\left(1-ŷᵢ\right)\right]\) (3) where N is the number of samples in the dataset, y i ​ is the ground truth label for sample i (1 for positive or real, 0 for negative or fake), and p i ​ is the predicted probability that sample ii belongs to the positive class (real) as determined by the discriminator. This loss guides the adversarial training process, improving the discriminator’s ability to distinguish between real and generated representations. Loss function: The core of this framework is the primary loss function, which integrates several distinct loss components. For our latent space, we include separate losses for fake and real representations, combined into an adversarial loss. Recognizing the inadequacy of adversarial loss alone, we supplement it with both decoder and encoder losses (see Fig. 1 -B-c). The composite formulation of the total loss function is: Total Loss = (α · Autoencoder Loss) + (β · Classifier Loss) + (γ · Adversarial Loss) Here, α, β, and γ are hyperparameters. This integration of loss terms facilitates precise classification and balances the critical tasks of discrimination, reconstruction, and feature extraction. This multifaceted approach enhances the overall performance and capabilities of our model. 3.8 Hyperparameter Selection The selection of hyperparameters, such as the learning rate, number of epochs, and dropout rates, was performed through a grid search approach combined with cross-validation. Specifically, we used a 5-fold cross-validation method on the training sets to tune the following key hyperparameters: Learning Rate: We experimented with learning rates in the range of 0.0001 to 0.001, ultimately selecting a value of 0.005, which provided the best balance between convergence speed and model performance without overfitting. Epochs: A total of 20 training epochs was chosen after evaluating the model’s performance across different values ranging from 10 to 30 epochs. We found that 20 epochs resulted in stable convergence without significant overfitting or underfitting. Batch Size: The optimal batch size was set to 32 after testing batch sizes of 16, 32, and 64, with 32 achieving the best trade-off between computational efficiency and model accuracy. Dropout Rate: To prevent overfitting, we applied a dropout rate of 0.3 after experimenting with values between 0.1 and 0.3, with 0.1 minimizing validation loss while maintaining training stability. These hyperparameters were optimized independently for each cell type, with results showing that this configuration achieved robust performance across different cell types and disease conditions. 3.9 Gene Ranking After analyzing and classifying disease and normal samples using our deep learning model, we focused on ranking genes to identify the most significant ones. By multiplying the weight matrices of the neural network layers, we obtained a final weight matrix with two dimensions: columns representing disease or control conditions, and rows representing genes. For the disease column, we sorted this weight matrix in ascending order and ranked the genes accordingly (Fig. 1 -B-d). This method allowed us to prioritize genes associated with the disease, offering valuable insights for further biological research. 3.10 Functional and Pathway Enrichment Analysis To explore the biological significance of the identified genes the top-ranked genes from each cell type and condition were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis using the Gprofiler tools. This allowed us to categorize the genes into relevant biological processes, cellular components, and molecular functions. Additionally, we employed STRING to construct protein-protein interaction (PPI) networks, helping to identify key interactions among the candidate genes. This analysis revealed critical pathways involved in immune responses and neuroinflammatory processes. We also used GProfiler to assess the enrichment of specific pathways and to identify potential biomarkers linked to both Covid-19, MS, and AD. These findings are integrated into our gene ranking strategy and discussed in detail in the results section (Fig. 1 -C). 4 Results In this section, we present the outcomes of our comprehensive analysis of the genetic links between severe Covid-19, MS, and AD using scRNA-seq data. We detail the model training process, evaluation metrics, identification of significant genes, pathway analysis, candidate biomarker detection, and comparison of biomarker specificity across different Covid-19 variants and severe flu. The goal is to elucidate the molecular mechanisms potentially connecting these diseases and to identify unique biomarkers that could serve as specific indicators for early disease detection and progression monitoring. In our experiments, we configured the hyperparameters as follows: β and γ were both set to 1, while α was set to 0.5. These settings were carefully selected to balance the contributions of the autoencoder, classifier, and adversarial losses within the total loss function. Specifically, assigning equal weights to β and γ ensured that both the classifier and adversarial losses were prioritized equally, promoting robust classification performance and effective adversarial training. The value of α was set to 0.5 to moderately weigh the autoencoder loss, maintaining an optimal trade-off between feature extraction and reconstruction accuracy. This configuration allowed the model to efficiently learn the underlying patterns in the scRNA-seq data, while effectively distinguishing between disease and control samples. By balancing these loss components, our model achieved a stable training process, minimizing overfitting and enhancing generalizability across different cell types and disease conditions. While the primary focus of our analysis centers on the genetic links between Covid-19 and MS, the investigation of Covid-19 and AD followed a similar approach. Both analyses utilized single-cell RNA sequencing (scRNA-seq) data and applied deep neural network models to identify relevant genetic interactions. Although we present the results for Covid-19 and AD in a more concise manner, the methodologies and analytical processes used for both disease comparisons were consistent, ensuring that the findings for each condition were derived from comparable techniques. 4.1 Model Training and Evaluation Across Cell Types The performance of the model was evaluated using several metrics, including accuracy, AUROC (Area Under the Receiver Operating Characteristic curve), and AUPR (Area Under the Precision-Recall curve), for each cell type (T-cells, B-cells, NK-cells, and monocytes) across the two disease conditions: Covid-19, and MS. The model demonstrated high classification accuracy and significant improvement in identifying disease-associated gene expression patterns compared to traditional methods. We trained our deep learning model separately on each cell type—T-cells, B-cells, NK-cells, and monocytes—using data from Covid-19 and MS conditions. For each disease, 20% of the raw data was reserved for validation, while the remaining 80% was used for data processing and gene selection as the training set. Table 2 summarizes the number of cells and selected genes for each cell type across the different conditions. This distribution highlights the variability in cell counts and gene selection between conditions, which is crucial for understanding the model's performance on diverse datasets. Figure 2 also illustrates the performance measures of the deep learning method for each cell type and disease. Table 2 The number of cells and selected genes for each cell type under different conditions Condition T-cell (cell/gene) B-cell (cell/gene) NK-cell (cell/gene) Monocyte (cell/gene) Covid-19 19153/251 4166/359 3854/310 9694/432 Multiple Sclerosis 21139/1199 4139/1112 5227/1249 10782/1410 This table provides a breakdown of cell counts and gene selection across T-cells, B-cells, NK-cells, and monocytes for Covid-19 and MS. 4.2 Identification of Significant Genes To identify the most significant genes for each cell type and disease condition, we applied our gene ranking method. This method involves multiplying the weight matrices obtained from the neural network layers, resulting in a 2 × n matrix, where 2 represents the disease and control conditions, and n is the number of significant genes identified for each cell type. Genes were ranked based on their weights in the disease column, with higher weights indicating greater relevance to the disease state. The ranked genes for each condition and cell type are presented in Supplementary Tables 1 through 12. These genes offer valuable insights into the potential molecular mechanisms underlying the diseases and highlight the genetic links between Covid-19, MS, AD, and neurodegenerative conditions. 4.3 Biological Relevance and Pathway Analysis Integrating gene ranking with pathway analysis allowed us to uncover the molecular mechanisms that potentially link Covid-19 and neurodegenerative diseases such as MS. By focusing on the top 100 ranked genes for each disease and cell type, we identified the most relevant genetic factors that may drive disease progression and interaction. This approach is particularly valuable for identifying subtle but crucial genetic interactions that might be overlooked in broader analyses. Several of the top-ranked genes were found to be non-coding RNAs (such as XIST, THUMPD3-AS1, LINC00513, MIF-AS1 , etc.), which play significant regulatory roles by indirectly influencing the expression of genes involved in critical cellular processes such as proliferation, migration, apoptosis, and immune responses. Their inclusion underscores the complexity of genetic regulation in these diseases and the need for detailed functional exploration (Supplementary Table 13). Following gene scoring, we selected the top 100 ranked genes for each disease and cell type and combined them, resulting in a set of 200 genes. The proteins encoded or regulated by these genes were analyzed using gProfiler 35 to identify enriched biological pathways, revealing several shared and distinct pathways across different cell types (Table 3 ) (Supplementary Fig. 1–4). In order to investigate more precisely, the Protein-Protein Interaction (PPI) network between the sets of genes in each cell type was drawn by STRING 36 (Supplementary Fig. 5–8). These pathways reflect roles in immune regulation, response to infection, and inflammatory processes, providing deeper insights into how Covid-19 may influence the progression of neurodegenerative diseases, particularly through immune-related mechanisms. Table 3 Pathway enrichment analysis identified for each cell type using GProfiler B-cell log10 (p) T-cell Log 10 (p) NK-cell log10 (p) Monocyte log10 (p) Interferon \(\:\gamma\:\) signaling -12.35 Immune response -7.69 Immune response -11.14 Positive regulation of immune response -9.6 Immune response -8.95 Regulation of reaction oxygen species metabolism -7.68 Interferon signaling -8.13 Regulation of vesicle-mediated transport -7.25 Programmed cell death -7.69 SARS-CoV2 signaling -5.88 Epstein-barr virus infection -7.11 Cytokine signaling in the immune system -6.12 SARS-CoV2 infections -6.23 Cell motility -5.77 T-cell activation -6.35 Regulation of receptor binding -5.45 TNF \(\:\alpha\:\) signaling pathway -5.65 Positive regulation of apoptosis -5.19 SARS-CoV2 signaling -5.88 Inflammatory response to antigenic stimulus -5.1 MAPK signaling pathway -5.16 TNF \(\:\alpha\:\) signaling pathway -4.65 Cellular response to cytokine stimuli -5.59 Regulation of protein kinase activity -4.95 Regulation of INF signaling -5.12 Cytokine signaling -4.64 Regulation of B-cell activation -4.45 Positive regulation of cell migration 4.63 This table lists the key pathways associated with the top 200 genes (combined top 100 genes from Covid-19 and MS) for each cell type—B-cells, T-cells, NK cells, and monocytes. "log10 (p)" is the p-value in log base 10. 4.4 Candidate Biomarker Detection To identify the most significant genes for each disease and cell type, we refined our analysis by narrowing down the initial set of 100 genes to a smaller, more focused subset. This step is crucial for effective biomarker discovery, as working with a large number of genes can be impractical and may dilute the specificity of the findings. The top-ranked genes were sequentially evaluated using classical machine learning algorithms, such as support vector machines and random forests, to calculate their F1 scores. Genes were added incrementally until the combined F1 score of these selected genes closely approximated the performance of our deep-learning model. For example, in the case of Covid-19 monocyte, an F1 score of approximately 0.98 was achieved with just 5 genes, XIST, HLA-DRB5, IGKV2-30, SULT1A1 , and RNF144B , nearly matching the deep learning model's F1 score of 0.989. This approach allowed us to efficiently pinpoint key genes while maintaining high predictive performance. The differential expression level of selected genes for Covid-19, MS, and AD in each cell type are illustrated in Fig. 3 . These genes represent potential biomarkers, showing a strong association with the respective diseases and highlighting their significance in disease mechanisms and progression. Among the selected genes, XIST and HLA-DRB5 stand out for their presence in multiple cell types and their association with both diseases. XIST has been identified as a hub lncRNA in MS 37 and is implicated in immune response regulation, particularly in females, due to its role in activating interferon-α production via a TLR7 -dependent mechanism 38 . Elevated expression of XIST (average log2FC of 12.79 in Covid-19 and 12.98 in MS) suggests its involvement in the immune response, particularly in the context of viral infections like Covid-19, where it may contribute to the observed female bias in disease severity 39 . HLA-DRB5 , part of the major histocompatibility complex (MHC) class II 40 , is crucial for presenting extracellular antigens to T-lymphocytes, thus enhancing immune responses 41 . Variations in HLA-DRB5 and other HLA class II loci have been shown to influence the severity of Covid-19 42 and the development of secondary progressive MS 43 . Our findings show increased expression of HLA-DRB5 (avg_log2FC of 1.44 in Covid-19 and 2.10 in MS), indicating its potential role in both diseases and highlighting its strong association with HLA-DRB1 in the gene and PPI networks. 4.5 Pathway and Protein-Protein Interaction Analysis To further explore the functional implications of these candidate genes, we performed pathway analysis and constructed protein-protein interaction (PPI) networks using STRING. These analyses revealed key interactions and pathways that contribute to disease progression in both Covid-19 and MS. For example, in B cells, the XIST (HNRNPU) gene can activate the MAPK pathway through DDX3X , leading to increased production of pro-inflammatory genes 44 . Figure 4 illustrates this mechanism, where XIST binds to MYD88 , activating several Toll-like receptors ( TLRs ), including TLR1, TLR2, TLR4, TLR5, and TLR6 45 . This activation triggers downstream pathways such as MKK4 and IKKα/β , ultimately enhancing the activity of transcription factors like JNK, FOS , and JUN , which leads to elevated expression of pro-inflammatory genes. Additionally, XIST activates endosomal TLRs ( TLR3, TLR7, TLR8, and TLR9 ), resulting in the activation of the NFκB pathway and subsequent release of inflammatory cytokines 46 . This process can exacerbate immune responses and contribute to CNS inflammation. Furthermore, DDX3X is recruited to viral replication sites, where it interacts with scaffold proteins of stress granules, including G3BP1, G3BP2, IGF2BP2, IGF2BP3 , and YBX1 47 . DDX3X also participates in antiviral signaling pathways by regulating the activity of TBK1, IKKε, TRAF3 , and IRF3 , ultimately leading to the production of type I interferon and modulation of the immune response 48 (Supplementary Fig. 9). Similarly, Fig. 5 demonstrates the role of HLA-DRB5 in the immune response. It shows how interferon production activates the JAK/STAT pathway, inducing IRF-1 , which then activates the CIITA gene 49 . This gene complex with other transcription factors ( ATF1/CREB, NF-Y , and RFX ) promotes the expression of MHC II genes 50 . This mechanism enhances the immune system's ability to present antigens and activate T-lymphocytes, potentially contributing to the inflammatory processes observed in both diseases 51 . In NK cells, MYL6B and MYO1F , which are involved in cell motility. These genes facilitate the migration of immune cells from the blood to other tissues, such as the brain, where they can exacerbate inflammation and contribute to neurodegeneration 52 . The increased expression of MYL6B and MYO1F in Covid-19 and MS further supports their role in the progression of these diseases 53 . These findings provide a comprehensive view of the molecular mechanisms linking Covid-19 and MS and underscore the potential of XIST , HLA-DRB5 , and MYO1F as candidate biomarkers for early disease detection. Further validation through in vitro and in vivo experiments is required to confirm their utility as blood-based markers for these diseases. Extending our analysis to AD, we aimed to determine whether similar immune regulatory mechanisms might be influencing its pathogenesis. Although AD, Covid-19, and MS are distinct conditions, they share common pathways of immune dysregulation and inflammation, which are known to be modulated by sex-specific factors. AD is the most common form of dementia, comprising 60–70% of all cases, and is increasingly recognized as a disorder involving both autoimmune and autoinflammatory mechanisms. In response to initial stimuli, amyloid-beta (Aβ) acts as an early reactive immunopeptide, triggering an innate immune cascade that leads to a chronic cycle of neuroinflammation and neuronal damage. Epidemiological data reveal a higher prevalence of AD in women, with a ratio of approximately 2:1 compared to men. This sex difference may be partly attributed to the age-related decline in estrogens post-menopause, along with incomplete X-chromosome inactivation that enhances the expression of immune-related genes in females. To explore potential overlaps between AD and the gene regulatory networks we identified in Covid-19 and MS, we analyzed an AD scRNA-seq dataset. While there was no direct overlap among the top-ranked genes, we identified key inflammatory genes in peripheral blood that are involved in pathways similar to those regulated by the Covid-19 and MS-associated genes. For instance, the genes DDX3X and XIST from Covid-19 were found to stimulate YBX1 in AD, while HLA-C in AD interacts with HLA-DQB1 and HLA-DRB5 from Covid-19 and MS, respectively 54 , 55 . These interactions are visualized in the STRING network (Fig. 6 ), which highlights the shared molecular pathways and interactions between these diseases. This suggests that common mechanisms of immune activation and inflammation may underlie these conditions, contributing to their pathogenesis. This Fig. provides a consolidated view of our findings by displaying the interconnected pathways and genes identified across Covid-19, MS, and AD, as detailed in Figs. 4 and 5 . This PPI network illustrates the roles of shared genes, such as XIST , DDX3X , and HLA-DRB5 , in immune regulation and inflammation, highlighting their potential as early biomarkers for neurodegenerative diseases. By mapping these genes across the conditions, Fig. 6 encapsulates the immune activation mechanisms contributing to disease progression and underscores the need for further exploration into shared pathways of immune dysregulation. Our findings indicate that genes such as YBX1 56 , DUSP1 57 , and HLA-C 58 could serve as potential biomarkers for early detection of AD due to their role in inflammation-induced neurodegeneration 59 . Overexpression of YBX1 in T, B, and NK cells enhances immune cell survival 60 , while DUSP1 prevents apoptosis 61 , potentially protecting neuronal cells 62 . Similarly, HLA-C facilitates antigen presentation, which could exacerbate neuroinflammation when overexpressed 63 . Given the significant roles these genes play in both systemic and CNS immune responses, further in vitro and in vivo validation is essential to confirm their potential as blood-based biomarkers for AD. 4.6 Comparison of Biomarker Specificity Across Covid-19 Variants and Severe Flu To assess whether the candidate biomarkers identified for severe Covid-19 and MS are unique to these conditions, we applied our methodology to an additional scRNA-seq dataset comprising samples from mild and asymptomatic Covid-19 cases, as well as severe flu cases. Using the same approach of integrating gene ranking with pathway analysis, we aimed to uncover the molecular mechanisms potentially linking these conditions to neurodegenerative diseases like MS. These datasets were the only publicly available PBMCs resources suitable for such a comparison, ensuring the robustness and transparency of our analysis. Our findings revealed that there was no overlap between the significant genes identified in severe Covid-19 and MS with those detected in asymptomatic Covid-19 or severe flu cases. This suggests that the biomarkers associated with severe Covid-19 and MS are distinct and not shared with less severe forms of Covid-19 or with severe flu. Consequently, these genetic markers may serve as specific indicators for severe Covid-19 and MS, distinguishing them from both milder Covid-19 cases and other respiratory conditions. The detailed results of this comparative analysis are presented in Supplementary Tables 14 through 21. 5 Discussion In our study, we identified several key genes and pathways significantly associated with severe Covid-19 MS and AD. Notably, the differential expression of genes such as XIST, HLA-DRB5, DDX3X, YBX1, DUSP1 , and HLA-C across various cell types suggests a potential link between these conditions and immune regulation. These findings highlight the role of non-coding RNAs and X-linked genes, indicating a complex interplay between genetic and environmental factors in disease progression. Given that XIST and DDX3X are involved in X-chromosome inactivation 64 and immune regulation 65 , their higher expression in females could enhance immune responses, potentially increasing susceptibility to autoimmune and neurodegenerative diseases like MS and AD while providing a protective effect against severe viral infections like Covid-19 66 . Sex has been shown to significantly influence immune responses at multiple levels—chromosomal, epigenetic, and hormonal—affecting disease onset, progression, and prognosis 67 . Females generally exhibit stronger inflammatory, antiviral, and humoral immune responses compared to males due to variations in both innate and adaptive immunity 68 , as evidenced by their higher phagocytic activity of neutrophils and macrophages, and more competent antigen-presenting cells (APCs). Conversely, males have a greater number of natural killers cells 69 . These immunological disparities are reflected in epidemiological data showing that males experience higher severity and fatality rates from Covid-19, potentially due to higher expression levels of SARS-CoV-2 entry receptors like ACE2 and TMPRSS2 . To further elucidate the interconnected mechanisms by which SARS-CoV-2 may influence CNS inflammation and immune responses in both Covid-19 and MS, we present Fig. 7 . This Fig. synthesizes findings from our pathway analyses and highlights the molecular cascade initiated by SARS-CoV-2 that impacts the CNS. Specifically, as shown in Fig. 4 , SARS-CoV-2 infiltrates the CNS by crossing the blood-brain barrier, where it activates microglia—the resident immune cells of the brain—triggering the release of inflammatory cytokines and activating T and B cells, thereby contributing to nerve damage. On the other hand, the virus also increases the expression of cell motility-related genes, including MYO1F and MYL6B , facilitating the migration of immune cells from the blood into brain tissue, worsening neuroinflammation and promoting neurodegeneration in conditions such as Covid-19, MS and AD. These findings align with previous data on XIST , HLA-DRB5 , and MYO1F, which were identified as potential biomarkers in the pathway analysis and are now underscored in Fig. 7 as pivotal players in CNS immune responses and disease progression. Together, this evidence emphasizes the need for further study of these genes as blood-based markers, offering insight into potential early detection strategies for both diseases. Overall, the sex-specific immune mechanisms we observed in Covid-19 and MS appear to extend to AD, underscoring the importance of personalized approaches in understanding and treating neurodegenerative diseases. Our study suggests that shared pathways of immune dysregulation and inflammation, influenced by sex and genetic factors, may underlie these conditions. Future research should aim to further elucidate these connections, potentially paving the way for targeted therapeutic interventions that take into account both genetic and sex-specific factors (Fig. 7 ) 6 Limitations of the study This study has some limitations that should be noted. We relied on publicly available scRNA-seq data, which may not fully represent all patient populations or disease conditions. The datasets were collected under different conditions, which might have introduced some variability, even though we addressed this during preprocessing. Additionally, since this is a computational study, experimental validation of the identified biomarkers is needed to confirm their biological relevance. Future research with larger and more diverse datasets will help strengthen and expand these findings. 7 Conclusion This study leveraged PBMC scRNA-seq to unravel the molecular connections between Covid-19, MS, and potentially AD. By analyzing gene expression at the single-cell level, we identified key genes such as XIST , DDX3X , and HLA-DRB5 , which are differentially expressed across various immune cell types in the context of Covid-19 and MS. These genes are involved in critical immune regulatory pathways, highlighting their role in modulating disease susceptibility and progression. Our findings suggest a potential link between Covid-19 and the onset or exacerbation of MS and AD, as both diseases share common inflammatory pathways and mechanisms of immune dysregulation. This connection is particularly relevant for individuals with a family history of MS or AD, as the expression of the identified candidate genes in PBMCs could serve as early biomarkers for assessing the risk of developing these neurodegenerative diseases following Covid-19 infection. Importantly, our analysis included both differentially expressed genes (DEGs) and non-DEGs, the latter proving significant even though they were often excluded in prior research. The candidate genes identified, including XIST , HLA-DRB5, MYO1F, YBX1, DUSP1 , and HLA-C , hold promise as biomarkers for the early detection of MS and AD. Evaluating their expression levels in blood samples could facilitate the identification of individuals at higher risk of progressing from Covid-19 to MS or developing AD, particularly in those with genetic predispositions. This approach offers a non-invasive and accessible means of early diagnosis, potentially enabling timely interventions that could mitigate disease progression. In summary, our study underscores the utility of PBMC scRNA-seq in uncovering the molecular links between Covid-19 and neurodegenerative diseases. The identified candidate genes lay the groundwork for developing blood-based diagnostic tools that can screen individuals with a family history of MS or AD, ultimately contributing to more personalized and preventive healthcare strategies. Future research should focus on validating these biomarkers in larger cohorts and exploring their clinical utility in early disease detection and risk assessment. 8 Theory and Calculation 8.1 Theory The theoretical foundation of this work is based on the intersection of infectious diseases and neurodegeneration, specifically the impact of severe Covid-19 on the immune system and its role in the progression of neurodegenerative diseases like MS and AD. Leveraging prior studies on immune dysregulation and inflammatory pathways, this work builds upon established knowledge of PBMC behavior and gene expression changes under pathological conditions. The study extends this theoretical framework by integrating novel insights into shared genetic and molecular mechanisms between Covid-19 and neurodegeneration, with a focus on identifying non-invasive biomarkers. 8.2 Calculation The practical implementation of the theory is achieved through a combination of advanced computational and statistical methods. These include: Deep Learning Models: Utilized autoencoders, classifiers, and adversarial training to process and analyze high-dimensional scRNA-seq data. Differential Gene Expression Analysis: Identified DEGs and non-DEGs, emphasizing their functional roles in immune and neurodegenerative processes. Pathway Enrichment and Network Analysis: Conducted pathway enrichment to reveal significant biological pathways and constructed protein-protein interaction networks to identify key regulatory nodes. Biomarker Prioritization: Prioritized blood biomarkers over brain biomarkers for early 9 Glossary PBMC (Peripheral Blood Mononuclear Cells): A type of blood cell that includes lymphocytes (T cells, B cells, and NK cells) and monocytes, used for analyzing immune responses. scRNA-seq (Single-Cell RNA Sequencing): A technique to study gene expression at the individual cell level, allowing for detailed molecular profiling. DEGs (Differentially Expressed Genes): Genes showing significant changes in expression levels between two conditions. Non-DEGs: Genes that do not show significant changes in expression but may still play subtle, critical roles in disease processes. Autoencoder: A type of artificial neural network used to learn efficient representations of data, particularly for dimensionality reduction. Adversarial Training: A machine learning approach where models are trained to be robust against small perturbations or adversarial inputs. Biomarker: A measurable indicator of a biological state or condition, used for diagnostics or monitoring disease progression. Pathway Enrichment Analysis: A computational method to identify biological pathways significantly impacted by a set of genes. Protein-Protein Interaction (PPI) Network: A graphical representation of interactions between proteins, used to identify key functional hubs in biological systems. Declarations Data Availability All transcriptomic datasets analyzed in this study were obtained from the Gene Expression Omnibus (GEO) repository. The following accession numbers were used: GSE166992, GSE138266, GSE181279 and GSE149689. All data are publicly available at https://www.ncbi.nlm.nih.gov/geo/. Code Availability The code for the deep learning model is accessible on GitHub repository [https://github.com/Shokoofeh3433/Deep-Learning-Model] Acknowledgments We would like to thank the School of Biological Sciences, Institute for Research in Fundamental Sciences (IPM) and Computing Center of IPM in performing parallel computing. Funding This study was conducted without any external funding support. Competing interests The authors report no competing interests. Conflict of Interest The authors declare no conflict of interest regarding the research, authorship, and publication of this article. Author Contribution Conceptualization: Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh. Data curation: Shokoofeh Ghiam. Formal analysis: Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor, Changiz Eslahchi. Investigation: Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh, Changiz Eslahchi. Methodology: Shokoofeh Ghiam, Changiz Eslahchi. Project administration: Asiyeh Mirzaei, Shokoofeh Ghiam. Resources: Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor. Software: Asiyeh Mirzaei, Mohammad Shirinpoor. Supervision: Shokoofeh Ghiam, Changiz Eslahchi. Validation: Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh, Mohammad Shirinpoor. Writing – original draft: Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor. Writing – review & editing: Pourya Naderi-Yeganeh, Shokoofeh Ghiam, Asiyeh Mirzaei, Mohammad Shirinpoor, Changiz Eslahchi. References Dhama K, Patel SK, Pathak M, et al. An update on SARS-CoV-2/COVID-19 with particular reference to its clinical pathology, pathogenesis, immunopathology and mitigation strategies. Travel Medicine and Infectious Disease. 2020; 37:101755. doi: 10.1016/j.tmaid.2020.101755 Camerlingo C. COVID-19 and neurodegenerative diseases. European Review. June 28, 2022. http://www.europeanreview.org/article/29093 Spudich S, Nath A. Nervous system consequences of COVID-19. Science. 2022;375(6578):267-269. doi:10.1126/science. abm2052 Sarwar S, Rogers S, Mohamed AS, et al. Multiple Sclerosis Following SARS-CoV-2 Infection: A Case Report and Literature Review. Cureus. Published online October 25, 2021. doi:10.7759/cureus.19036 MacDougall M, El-Hajj Sleiman J, Beauchemin P, Rangachari M. SARS-CoV-2 and Multiple Sclerosis: Potential for Disease Exacerbation. Frontiers in Immunology. 2022;13. doi:10.3389/fimmu.2022.871276 Bassetti M, Vena A, Giacobbe DR. The novel Chinese coronavirus (2019‐nCoV) infections: Challenges for fighting the storm. European Journal of Clinical Investigation. 2020;50(3). doi:10.1111/eci.13209 Ahmed SF, Quadeer AA, McKay MR. Preliminary Identification of Potential Vaccine Targets for the COVID-19 Coronavirus (SARS-CoV-2) Based on SARS-CoV Immunological Studies. Viruses. 2020;12(3):254. doi:10.3390/v12030254 Hoang HD, Naeli P, Alain T, Jafarnejad SM. Mechanisms of impairment of interferon production by SARS-CoV-2. Biochemical Society Transactions. 2023;51(3):1047-1056. doi:10.1042/bst20221037 Yang L, Xie X, Tu Z, Fu J, Xu D, Zhou Y. The signal pathways and treatment of cytokine storm in COVID-19. Signal Transduction and Targeted Therapy. 2021;6(1). doi:10.1038/s41392-021-00679-0 Zang X, Chen S, Zhu J, Ma J, Zhai Y. The Emerging Role of Central and Peripheral Immune Systems in Neurodegenerative Diseases. Frontiers in Aging Neuroscience. 2022;14. doi:10.3389/fnagi.2022.872134 Hui DS, I Azhar E, Madani TA, et al. The continuing 2019-nCoV epidemic threat of novel coronaviruses to global health — The latest 2019 novel coronavirus outbreak in Wuhan, China. International Journal of Infectious Diseases. 2020; 91:264-266. doi: 10.1016/j.ijid.2020.01.009 Ahmad T, Khan M, Haroon, et al. COVID-19: Zoonotic aspects. Travel Medicine and Infectious Disease. 2020; 36:101607. doi: 10.1016/j.tmaid.2020.101607 Bjornevik K, Münz C, Cohen JI, Ascherio A. Epstein–Barr virus as a leading cause of multiple sclerosis: mechanisms and implications. Nature Reviews Neurology. Published online February 9, 2023. doi:10.1038/s41582-023-00775-5 Balasa R, Barcutean L, Mosora O, Manu D. Reviewing the Significance of Blood–Brain Barrier Disruption in Multiple Sclerosis Pathology and Treatment. International Journal of Molecular Sciences. 2021;22(16):8370. doi:10.3390/ijms22168370 Park HJ, Jung H. Neuro-immune interactions at single-cell resolution in neurodevelopmental, infectious, and neurodegenerative diseases. Animal Cells and Systems. 2022;26(4):137-147. doi:10.1080/19768354.2022.2110937 Kumar D, Jahan S, Khan A, et al. Neurological Manifestation of SARS-CoV-2 Induced Inflammation and Possible Therapeutic Strategies Against COVID-19. Molecular Neurobiology. Published online March 14, 2021. doi:10.1007/s12035-021-02318-9 Li X, Wang CY. From bulk, single-cell to spatial RNA sequencing. International Journal of Oral Science. 2021;13(1). doi:10.1038/s41368-021-00146-0 Al-Tashi Q, Saad MB, Muneer A, et al. Machine Learning Models for the Identification of Prognostic and Predictive Cancer Biomarkers: A Systematic Review. International Journal of Molecular Sciences. 2023;24(9):7781. doi:10.3390/ijms24097781 Biernacki T, Kokas Z, Sandi D, et al. Emerging Biomarkers of Multiple Sclerosis in the Blood and the CSF: A Focus on Neurofilaments and Therapeutic Considerations. International Journal of Molecular Sciences. 2022;23(6):3383. doi:10.3390/ijms23063383 J. Huang, M. Khademi, L. Fugger, Ö. Lindhe, L. Novakova, M. Axelsson, C. Malmeström, C. Constantinescu, J. Lycke, F. Piehl, T. Olsson, I. Kockum, Inflammation-related plasma and CSF biomarkers for multiple sclerosis, Proc. Natl. Acad. Sci. U.S.A.117 (23) 12952-12960, https://doi.org/10.1073/pnas.1912839117 (2020). Thompson EA, Cascino K, Ordonez AA, et al. Metabolic programs define dysfunctional immune responses in severe COVID-19 patients. Cell Reports. 2021;34(11):108863. doi: 10.1016/j.celrep.2021.108863 Talebi H, Ghiam S, Mirzaei Koli A, Naderi Yeganeh P, Eslahchi C. GeneDX-PBMC: An adversarial autoencoder framework for unlocking Alzheimer’s disease biomarkers using blood single-cell RNA sequencing data. Comput Biol Med . 2025; 192:110283. doi: 10.1016/j.compbiomed.2025.110283 Schafflick D, Xu CA, Hartlehnert M, et al. Integrated single cell analysis of blood and cerebrospinal fluid leukocytes in multiple sclerosis. Nature Communications. 2020;11(1). doi:10.1038/s41467-019-14118-w Xu H, Jia J. Single-Cell RNA Sequencing of Peripheral Blood Reveals Immune Cell Signatures in Alzheimer’s Disease. Frontiers in Immunology. 2021;12. doi:10.3389/fimmu.2021.645666 Barrett T, Wilhite SE, Ledoux P, et al. NCBI GEO: archive for functional genomics data sets—update. Nucleic Acids Research. 2012;41(D1): D991-D995. doi:10.1093/nar/gks1193 Lee JS, Park S, Jeong HW, et al. Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19. Science Immunology. 2020;5(49). doi:10.1126/sciimmunol. abd1554 Hao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573-3587.e29. doi: 10.1016/j.cell.2021.04.048 Hao et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nature Biotechnology (2023) [Seurat V5] R: A Language and Environment for Statistical Computing: Reference Index.; 2010. Aran D, Looney AP, Liu L, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nature Immunology. 2019;20(2):163-172. doi:10.1038/s41590-018-0276-y Chintala AP Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, Soumith. PyTorch: An imperative style, high-performance deep learning library. NeurIPS. 2019;32(1). Creswell A, White T, Dumoulin V, Arulkumaran K, Sengupta B, Bharath AA. Generative Adversarial Networks: An Overview. IEEE Signal Processing Magazine. 2018;35(1):53-65. doi:10.1109/msp.2017.2765202 Szklarczyk D, Kirsch R, Koutrouli M, et al. The STRING database in 2023: protein–protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Research. 2022;51(D1): D638-D646. doi:10.1093/nar/gkac1000 Kolberg L, Raudvere U, Kuzmin I, Adler P, Vilo J, Peterson H. g: Profiler—interoperable web service for functional enrichment analysis and gene identifier mapping (2023 update). Nucleic Acids Research. 2023;51(W1): W207-W212. doi:10.1093/nar/gkad347 Ding Y, Li T, Yan X, et al. Identification of hub lncRNA ceRNAs in multiple sclerosis based on ceRNA mechanisms. Molecular Genetics and Genomics. 2021;296(2):423-435. doi:10.1007/s00438-020-01750-1 Souyris M, Cenac C, Azar P, et al. TLR7 escapes X chromosome inactivation in immune cells. Science Immunology. 2018;3(19). doi:10.1126/sciimmunol. aap8855 Yu B, Qi Y, Li R, Shi Q, Satpathy AT, Chang HY. B cell-specific XIST complex enforces X-inactivation and restrains atypical B cells. Cell. 2021;184(7):1790-1803.e17. doi: 10.1016/j.cell.2021.02.015 Gumkowska-Sroka, O., Kotyla, K., & Kotyla, P. (2024). Immunogenetics of Systemic Sclerosis. Genes, 15(5), 586. https://doi.org/10.3390/genes15050586 Medhasi S, Chantratita N. Human Leukocyte Antigen (HLA) System: Genetics and Association with Bacterial and Viral Infections. Journal of Immunology Research. 2022; 2022:1-15. doi:10.1155/2022/9710376 Hoseinnezhad T, Soltani N, Ziarati S, Behboudi E, Mousavi MJ. The role of HLA genetic variants in COVID‐19 susceptibility, severity, and mortality: A global review. Journal of Clinical Laboratory Analysis. 2024;38(1-2). doi:10.1002/jcla.25005 Prapas P, Anagnostouli M. Macrophages and HLA-Class II Alleles in Multiple Sclerosis: Insights in Therapeutic Dynamics. International Journal of Molecular Sciences. 2024;25(13). doi:10.3390/ijms25137354 Kovarik P, Bestehorn A and Fesselet J (2021) Conceptual Advances in Control of Inflammation by the RNA-Binding Protein Tristetraprolin. Front. Immunol. 12:751313. doi: 10.3389/fimmu.2021.751313 Asami J, Shimizu T. Structural and functional understanding of the toll‐like receptors. Protein Science. 2021;30(4):761-772. doi:10.1002/pro.4043 Bost C, Arleevskaya MI, Brooks WH, Plaza S, Guery JC, Renaudineau Y. Long non-coding RNA Xist contribution in systemic lupus erythematosus and rheumatoid arthritis. Clinical Immunology. 2022; 236:108937. doi: 10.1016/j.clim.2022.108937 Ciccosanti F, Di Rienzo M, Romagnoli A, et al. Proteomic analysis identifies the RNA helicase DDX3X as a host target against SARS-CoV-2 infection. Antiviral Research. 2021; 190:105064. doi: 10.1016/j.antiviral.2021.105064 Hernández-Díaz T, Valiente-Echeverría F, Soto-Rifo R. RNA Helicase DDX3: A Double-Edged Sword for Viral Replication and Immune Signaling. Microorganisms. 2021;9(6):1206. doi:10.3390/microorganisms9061206 Tong J, Zhang W, Chen Y, Yuan Q, Qin N-N and Qu G (2022) The Emerging Role of RNA Modifications in the Regulation of Antiviral Innate Immunity. Front. Microbiol . 13:845625. doi: 10.3389/fmicb.2022.845625 Mays Mohammed Salih, Chi G Weindel, Eric Malekos, Lisa Sudek, Sol Katzman, Cory J Mabry, Aja K Coleman, Sikandar Azam, Robert Watson, Kristin Patrick, Susan Carpenter, bioRxiv 2023.10.12.562050; doi: https://doi.org/10.1101/2023.10.12.562050 Renata Mezyk-Kopec, Lambert Potin, J. Emiliano Gomez Medellin, Calixto M. Salles, Melody A. Swartz; TGF-β Signaling Prevents MHC Class II–Expressing Lymphatic Endothelial Cells from Reactivating Human Allogenic Memory CD4+ T Cells. J Immunol 1 September 2023; 211 (5): 782–790. https://doi.org/10.4049/jimmunol.2200216 Matozo T, Kogachi L, de Alencar BC. Myosin motors on the pathway of viral infections. Cytoskeleton. 2022;79(6-8):41-63. doi:10.1002/cm.21718 Matozo T, Kogachi L, de Alencar BC. Myosin motors on the pathway of viral infections. Cytoskeleton. 2022;79(6-8):41-63. doi:10.1002/cm.21718 Ghiam S, Eslahchi C, Shahpasand K, Habibi-Rezaei M, Gharaghani S. Exploring the role of non-coding RNAs as potential candidate biomarkers in the cross-talk between diabetes mellitus and Alzheimer’s disease. Frontiers in Aging Neuroscience. 2022;14. doi:10.3389/fnagi.2022.955461 Ghiam S, Eslahchi C, Shahpasand K, Habibi-Rezaei M, Gharaghani S. Identification of repurposed drugs targeting significant long non-coding RNAs in the cross-talk between diabetes mellitus and Alzheimer’s disease. Scientific Reports. 2022;12(1). doi:10.1038/s41598-022-22822-9 Kloetgen A, Duggimpudi S, Schuschel K, et al. YBX1 Indirectly Targets Heterochromatin-Repressed Inflammatory Response-Related Apoptosis Genes through Regulating CBX5 mRNA. International Journal of Molecular Sciences. 2020;21(12). doi:10.3390/ijms21124453 Pan AL, Audrain M, Sakakibara E, et al. Dual-Specificity Protein Phosphatase 4 (DUSP4) Overexpression Improves Learning Behavior Selectively in Female 5xFAD Mice, and Reduces β-Amyloid Load in Males and Females. Cells. 2022;11(23). doi:10.3390/cells11233880 Cătană CS, Marta MM, Văleanu M, Dican L, Crișan CA. Human Leukocyte Antigen and microRNAs as Key Orchestrators of Mild Cognitive Impairment and Alzheimer’s Disease: A Systematic Review. International Journal of Molecular Sciences. 2024;25(15). doi:10.3390/ijms25158544 Hayden AN, Brandel KL, Merlau PR, et al. Behavioral Screening of Conserved RNA-Binding Proteins Reveals CEY-1/YBX RNA-Binding Protein Dysfunction Leads to Impairments in Memory and Cognition. Cold Spring Harbor Laboratory; 2024. Accessed October 15, 2024. http://dx.doi.org/10.1101/2024.01.05.574402 Shao LL, Gao MM, Gong JX, Yang LY. DUSP1 regulates hippocampal damage in epilepsy rats via ERK1/2 pathway. Journal of Chemical Neuroanatomy. 2021; 118:102032. doi: 10.1016/j.jchemneu.2021.102032 Yang J, Sun L, Han J, Zheng W, Peng W. DUSP1/MKP-1 regulates proliferation and apoptosis in keratinocytes through the ERK/Elk-1/Egr-1 signaling pathway. Life Sciences. 2019; 223:47-53. doi: 10.1016/j.lfs.2019.03.018 Pérez-Sen R, Queipo MJ, Gil-Redondo JC, et al. Dual-Specificity Phosphatase Regulation in Neurons and Glial Cells. International Journal of Molecular Sciences. 2019;20(8):1999. doi:10.3390/ijms20081999 Gross CC, Meyer C, Bhatia U, et al. CD8+ T cell-mediated endotheliopathy is a targetable mechanism of neuro-inflammation in Susac syndrome. Nature Communications. 2019;10(1). doi:10.1038/s41467-019-13593-5 Fang H, Disteche CM, Berletch JB. X Inactivation and Escape: Epigenetic and Structural Features. Frontiers in Cell and Developmental Biology. 2019;7. doi:10.3389/fcell.2019.00219 Ryan CS and Schröder M (2022) The human DEAD-box helicase DDX3X as a regulator of mRNA translation. Front. Cell Dev. Biol. 10:1033684. doi: 10.3389/fcell.2022.1033684 Chlamydas S, Papavassiliou AG, Piperi C. Epigenetic mechanisms regulating COVID-19 infection. Epigenetics. 2020;16(3):263-270. doi:10.1080/15592294.2020.1796896 Dodd KC and Menon M (2022) Sex bias in lymphocytes: Implications for autoimmune diseases. Front. Immunol. 13:945762. doi: 10.3389/fimmu.2022.945762 Takahashi T, Iwasaki A. Sex differences in immune responses. Science . 2021;371(6527):347-348. doi:10.1126/science. abe7199 Jacobsen H and Klein SL (2021) Sex Differences in Immunity to Viral Infections. Front. Immunol. 12:720952. doi: 10.3389/fimmu.2021.720952 Kucuksezer UC, Aktas Cetin E, Esen F, Tahrali I, Akdeniz N, Gelmez MY, Deniz G. The Role of Natural Killer Cells in Autoimmune Diseases. Frontiers in Immunology . 2021; 12:622306. doi:10.3389/fimmu.2021.622306 Forsblom E, Silén S, Kortela E, et al. Male predominance in disease severity and mortality in a low Covid-19 epidemic and low case-fatality area – a population-based registry study. Infectious Diseases. 2021;53(10):789-799. doi:10.1080/23744235.2021.1936157 Additional Declarations No competing interests reported. Supplementary Files SupplementaryFigurese1.png SupplementaryFigurese2.png SupplementaryFigurese3.png SupplementaryFigurese4.png SupplementaryFigurese5.png SupplementaryFigurese6.png SupplementaryFigurese7.png SupplementaryFigurese8.png SupplementaryFigurese9.png Supplementary.docx Supplementarytable1.xlsx Supplementarytable2.xlsx Supplementarytable3.xlsx Supplementarytable4.xlsx Supplementarytable5.xlsx Supplementarytable6.xlsx Supplementarytable7.xlsx Supplementarytable8.xlsx Supplementarytable9.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7342199","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":524683056,"identity":"35586eb1-ade5-44ca-81a4-e75bab0eb219","order_by":0,"name":"Asiyeh Mirzaei Koli","email":"","orcid":"","institution":"Shahid-Beheshti University","correspondingAuthor":false,"prefix":"","firstName":"Asiyeh","middleName":"Mirzaei","lastName":"Koli","suffix":""},{"id":524683057,"identity":"2e6dfcd8-081b-44ee-a401-d3033cb22694","order_by":1,"name":"Shokoofeh Ghiam","email":"","orcid":"","institution":"Institute for Research in Fundamental Sciences (IPM)","correspondingAuthor":false,"prefix":"","firstName":"Shokoofeh","middleName":"","lastName":"Ghiam","suffix":""},{"id":524683058,"identity":"9e11d23c-b071-46b0-80d3-a4f74b78825c","order_by":2,"name":"Mohammad Shirinpoor Kharf","email":"","orcid":"","institution":"Shahid-Beheshti University","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Shirinpoor","lastName":"Kharf","suffix":""},{"id":524683059,"identity":"1ee55df9-4d5c-4e37-b050-5ed3671a6618","order_by":3,"name":"Pourya Naderi Yeganeh","email":"","orcid":"","institution":"Beth Israel Deaconess Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Pourya","middleName":"Naderi","lastName":"Yeganeh","suffix":""},{"id":524683060,"identity":"3bf950f2-5a1d-4a68-b18f-5a7df2b68eb2","order_by":4,"name":"Changiz Eslahchi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIie3RsWrCQBzH8d8R0OVC1j8E7Cv8xbHgs9zh4KYtXSwdjATS0TXgSzg6Jtzgcg8Q0EEfIUtJttqUToVTN4f7Dv/h4MP/Dwf4fI8Y/YwXKJBYnUTy+1hcJ9yRlO8l6NEfcRZtEnFueDyLNmn23u4Mos9CmIVrybEIRpInb3Qss0NoDcgqlNZBmOZfMTjQW9LZQWQGqIDSdSCT6rcNLzvy2l7I0w2kB8mmIwgvhK8RqlQQS97rvNJpHNqpHFqdOEmUK1E3iw+9zqfnut09DwZ7Y2oX+ZcEbvodn8/n87n6BjNLTvwpgJZtAAAAAElFTkSuQmCC","orcid":"","institution":"Shahid-Beheshti University","correspondingAuthor":true,"prefix":"","firstName":"Changiz","middleName":"","lastName":"Eslahchi","suffix":""}],"badges":[],"createdAt":"2025-08-11 04:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7342199/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7342199/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93013709,"identity":"d8a08a90-870b-4420-867c-15957c1e0c4e","added_by":"auto","created_at":"2025-10-08 07:28:27","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1844110,"visible":true,"origin":"","legend":"","description":"","filename":"ScientificReports.docx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/fda6c2a63fcc864f4bbb4f9e.docx"},{"id":93013708,"identity":"3e4901fa-37ee-425b-92d5-e08a7dcdef27","added_by":"auto","created_at":"2025-10-08 07:28:27","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7441,"visible":true,"origin":"","legend":"","description":"","filename":"71130d7117be43738c5e3f303cf253ca.json","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/08319a5653785cf7f11c66e2.json"},{"id":93011912,"identity":"bd8153ce-4ad6-475b-bdf9-ed3a7f044bf9","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":80551,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese1.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/c1b67c5216f20f01f483d5b6.png"},{"id":93011905,"identity":"8cfa5c68-6f52-44c4-8b50-afaad8b78ed3","added_by":"auto","created_at":"2025-10-08 07:20:27","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88905,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese2.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/d9652920f8046d6bc3ca3766.png"},{"id":93013710,"identity":"e22c7b12-5819-4617-bc9e-20f2ad2e2293","added_by":"auto","created_at":"2025-10-08 07:28:27","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":87913,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese3.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/c23787fe78bb644f4938c429.png"},{"id":93014851,"identity":"f3fbc812-5900-4e9c-8964-7786839d29f8","added_by":"auto","created_at":"2025-10-08 07:44:27","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":87706,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese4.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/acc970f222f7ffa4dc90d5ba.png"},{"id":93011915,"identity":"2f1fb392-6129-4864-8593-f99ce6d2cf44","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":458326,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese5.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/327a5cb48d0abdf6e1325897.png"},{"id":93014100,"identity":"8b6b7e89-040a-42ed-9e20-e7dc943198d9","added_by":"auto","created_at":"2025-10-08 07:36:28","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3438630,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese6.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/bd914880c8049cac5d68ac13.png"},{"id":93011919,"identity":"bf85590c-4369-4937-9d14-dde79239713b","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":441951,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese7.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/6250a38c4b814a96d098d58b.png"},{"id":93011929,"identity":"d8078f5d-2f6e-46a4-ae8c-cd1454d3d220","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":441951,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese8.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/b0e5f561f5bc3d374bb05aa3.png"},{"id":93014852,"identity":"a0340d77-525b-4fb0-bd52-970a6a71bdc6","added_by":"auto","created_at":"2025-10-08 07:44:28","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15423,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese9.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/0ebca7197081784b329a21a4.png"},{"id":93011944,"identity":"7525048b-dbcc-4a56-811f-c5024e021555","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6297286,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/e625346dc76e3cceb82b3e1f.docx"},{"id":93011951,"identity":"44da28de-5add-4ad9-a5c0-3741a9fb3e93","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"xlsx","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27404,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/34ea87e9b959c92cb54e426f.xlsx"},{"id":93014101,"identity":"783699b4-7abf-4f27-bcc7-d7dd8cfa80a6","added_by":"auto","created_at":"2025-10-08 07:36:28","extension":"xlsx","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45641,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/5c36aac8b00ed3cc0a98bcbe.xlsx"},{"id":93013721,"identity":"4003535a-dbd3-46f2-b0c8-aa0885a35b67","added_by":"auto","created_at":"2025-10-08 07:28:28","extension":"xlsx","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21384,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/6aab213eda13a6174463b9b7.xlsx"},{"id":93013725,"identity":"2dd07be9-59f2-4369-8734-57668f004b26","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"xlsx","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62659,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/0d8b65c144e94a2198109f0f.xlsx"},{"id":93013716,"identity":"a504d43b-c8c2-40e8-96c5-372d0e9ffe86","added_by":"auto","created_at":"2025-10-08 07:28:28","extension":"xlsx","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":115124,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/d5c86f679af7f85a6ea7d545.xlsx"},{"id":93014104,"identity":"e793b788-74a6-4544-a68f-7d68ded95220","added_by":"auto","created_at":"2025-10-08 07:36:29","extension":"xlsx","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46976,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/3973959e0aca98ee1e65d74a.xlsx"},{"id":93011934,"identity":"b5f6439b-3556-44b9-a473-1749a27dcaa7","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"xlsx","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":113118,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/4e5bb4705527261721dec9ac.xlsx"},{"id":93014106,"identity":"33ce3da3-1254-4e16-88a5-f3fd746d6b5e","added_by":"auto","created_at":"2025-10-08 07:36:29","extension":"xlsx","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":203878,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/b2f7c7e26d134df21d157647.xlsx"},{"id":93011943,"identity":"05383172-10ef-4d8f-a74a-d86cd127a50d","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"xlsx","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17205,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/f9b8c09078e0278b10545278.xlsx"},{"id":93011963,"identity":"adfba24f-bb9e-4c51-99bc-ec8f7c6f7cef","added_by":"auto","created_at":"2025-10-08 07:20:30","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":192120,"visible":true,"origin":"","legend":"","description":"","filename":"71130d7117be43738c5e3f303cf253ca1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/fb8c878ec4c1df9af20d84ca.xml"},{"id":93011959,"identity":"83ebebf2-28af-4727-89f5-e3c03c2ab0ee","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"jpeg","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":227705,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/1ea1e76691c212cd32c52b91.jpeg"},{"id":93014103,"identity":"0721ccaf-e65b-4df0-aa3d-0848e7adadc7","added_by":"auto","created_at":"2025-10-08 07:36:29","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":254967,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/67763c1fe88a2cfc91cadbc9.png"},{"id":93013734,"identity":"fb26eccb-b4b5-4742-9fe2-8dab687355cf","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":473907,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/df205a58c73535d049ba903e.png"},{"id":93011952,"identity":"156c1a6e-5774-449d-bc14-a62ffd2c8c47","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":79924,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/fbacdf3e5783224e398d8630.png"},{"id":93011940,"identity":"85a10150-59c8-4421-a361-34bb49f74aaf","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100561,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/81a63f1d0e212d4eeeed2986.png"},{"id":93013724,"identity":"758eb128-d479-416e-83ed-5b367df1ef27","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":272261,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/76362066494bfcf4ce8d4813.png"},{"id":93011941,"identity":"0a22d36a-d862-4d62-b170-75a68773053f","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":372277,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/e1d60b1e2578ec2a95edc954.png"},{"id":93011960,"identity":"27344556-2e3f-4514-9a6c-82db1775fa88","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126358,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/d2d4ed3624146767b1342556.png"},{"id":93014105,"identity":"9c34b97b-9efe-4d6a-aa67-0c94b02fa669","added_by":"auto","created_at":"2025-10-08 07:36:29","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50161,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/e42289c8143a36d093f98fd8.png"},{"id":93011946,"identity":"909c5088-8d81-457c-8ed9-a0fb56f857c3","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97824,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/23e6e63bb84cd6b39dbe3852.png"},{"id":93013723,"identity":"8f127138-24eb-447e-97f9-b40f29a62c08","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37858,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/39ce8ae9fc735e8855c288d2.png"},{"id":93011955,"identity":"02f5314c-37ba-4123-9b0f-13f7b5e6fb5e","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35536,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/7b5be03fa24877dffe19a9bd.png"},{"id":93011957,"identity":"03f54efd-20f0-4c13-9c98-c7cef809ada4","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52365,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/21b4c04175a8d21c4c2ac5be.png"},{"id":93013732,"identity":"e7af513e-5c17-4c3b-afd6-bcf503bea174","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98162,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/af16e9d1582d4ac8df856ec0.png"},{"id":93011962,"identity":"6905f92b-b57e-4cf1-a1fe-785b024e3fa1","added_by":"auto","created_at":"2025-10-08 07:20:30","extension":"xml","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":190770,"visible":true,"origin":"","legend":"","description":"","filename":"71130d7117be43738c5e3f303cf253ca1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/885e3b8cc315cd4874ef1243.xml"},{"id":93013729,"identity":"5cafc257-b44e-4005-86ea-3008c9c0806f","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"html","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":213197,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/023820cd77741b823550b76c.html"},{"id":93014096,"identity":"fe011a81-ff80-4502-8f71-afd16a75d80b","added_by":"auto","created_at":"2025-10-08 07:36:27","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":341322,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe workflow of the study. (A)\u003c/strong\u003e Single-cell RNA-seq Analysis: Gene expression datasets from the Gene Expression Omnibus (GEO) related to AD, MS, and Covid-19 are processed. After quality control and batch effect removal, cell clustering was performed and genes involved in the pathogenesis of each disease were identified. \u003cstrong\u003e(B)\u003c/strong\u003e Computational Model: A deep learning-based framework is developed to model the relationships between identified genes. Genes are ranked based on the multiplication of weight matrices. \u003cstrong\u003e(C)\u003c/strong\u003e Candidate Biomarker Prediction: Identified candidate biomarkers are subjected to further validation using tools like STRING for protein-protein interaction networks and Gprofiler for functional enrichment analysis. A K-nearest neighbors (KNN) approach is applied to improve the gene ranking for better biomarker identification.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/2e810f942bd9cc8bcd1660c4.jpeg"},{"id":93011909,"identity":"3ac6e4c0-a83c-4767-a4c2-555a9fb259b7","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":217979,"visible":true,"origin":"","legend":"\u003cp\u003ePerformance metrics of the model for each cell type and condition, evaluated using training data (P: Precision, R: Recall, F1: F1-score) and two different validation sets (val.1 and val.2). The Fig. shows the precision, recall, and F1-score for each cell type across different diseases, indicating the model's effectiveness in classifying normal and diseased samples.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/038a37bff01196cbfc952629.png"},{"id":93011920,"identity":"09f9bed9-beae-4ed7-9b66-a91f0d83d3e8","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":437551,"visible":true,"origin":"","legend":"\u003cp\u003eThis Fig. presents the top-ranked gene candidates for Covid-19, MS, and AD across various cell types. These genes were selected based on their performance in classical machine-learning algorithms, highlighting their potential as biomarkers for disease detection and progression monitoring. Notably, some genes exhibit low log2 fold changes (log2FC).\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/c5f8f9011cd9a07cd5169ddb.png"},{"id":93011901,"identity":"4bbcf7ea-1e5e-4fc7-b266-e5fecb3b3387","added_by":"auto","created_at":"2025-10-08 07:20:27","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":67005,"visible":true,"origin":"","legend":"\u003cp\u003eThis Fig. shows mechanisms through which \u003cem\u003eXIST\u003c/em\u003e stimulates pro-inflammatory gene production. \u003cstrong\u003e(a)\u003c/strong\u003e Pathway 1: \u003cem\u003eXIST\u003c/em\u003e binds to \u003cem\u003eMYD88\u003c/em\u003e, activating \u003cem\u003eTLR1\u003c/em\u003e, \u003cem\u003eTLR2\u003c/em\u003e, \u003cem\u003eTLR4\u003c/em\u003e, \u003cem\u003eTLR5\u003c/em\u003e, and \u003cem\u003eTLR6\u003c/em\u003e, leading to the \u003cem\u003eMKK4\u003c/em\u003e and \u003cem\u003eIKKα/β\u003c/em\u003e pathways. This results in the phosphorylation of \u003cem\u003eIRAK4\u003c/em\u003e and \u003cem\u003eIRAK1\u003c/em\u003e, which enhances transcription factors \u003cem\u003eJNK\u003c/em\u003e, \u003cem\u003eFOS\u003c/em\u003e, and \u003cem\u003eJUN\u003c/em\u003e, culminating in increased pro-inflammatory gene expression in the nucleus. \u003cstrong\u003e(b)\u003c/strong\u003e Pathway 2: \u003cem\u003eXIST\u003c/em\u003eactivates endosomal \u003cem\u003eTLRs\u003c/em\u003e (\u003cem\u003eTLR3, TLR7, TLR8, TLR9\u003c/em\u003e) through \u003cem\u003eTIRAP/MYD88\u003c/em\u003e and \u003cem\u003eIRAK4\u003c/em\u003e, inducing \u003cem\u003eTAK1\u003c/em\u003e. \u003cem\u003eTAK1\u003c/em\u003e activates the \u003cem\u003eNFκB\u003c/em\u003e pathway via the \u003cem\u003eIKK\u003c/em\u003epathway and \u003cem\u003eMAPK\u003c/em\u003e pathway transcription factors (\u003cem\u003eFOS, JUN\u003c/em\u003e), promoting pro-inflammatory gene expression in the nucleus. \u003cstrong\u003e(c)\u003c/strong\u003e Resulting inflammatory cytokines and interferon are released, recruiting immune cells to damaged tissues, such as the brain. This leads to inflammation and damage to the CNS, exacerbating the inflammatory process. The genes highlighted in yellow were observed in our results.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/40b3a22ce2da12168b44f18c.png"},{"id":93011906,"identity":"40d16ea3-cde4-44c7-9148-2cb2d2e100db","added_by":"auto","created_at":"2025-10-08 07:20:27","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":83618,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eINF\u003c/em\u003e leads to activation of the \u003cem\u003eJAK/STAT\u003c/em\u003e pathway which causes the induction of the \u003cem\u003eIRF-1 gene\u003c/em\u003e in the peripheral blood. Then \u003cem\u003eIRF-1\u003c/em\u003e binds to ISRE and causes the activation of the \u003cem\u003eCIITA\u003c/em\u003egene. The CIITA gene migrates into the nucleus, where it forms a complex with \u003cem\u003eATF1/CREB\u003c/em\u003e, \u003cem\u003eNF-Y\u003c/em\u003e and \u003cem\u003eRFX\u003c/em\u003e genes and causes the expression of \u003cem\u003eMHC II\u003c/em\u003e genes (fig. 3a). In addition, the \u003cem\u003eIRF-1\u003c/em\u003egene can cause the transactivation of the \u003cem\u003eNLRC5\u003c/em\u003egene. \u003cem\u003eNLRC5\u003c/em\u003e forms a complex with \u003cem\u003eATF1/CREB\u003c/em\u003e, \u003cem\u003eNF-Y\u003c/em\u003e and \u003cem\u003eRFX\u003c/em\u003e genes in the nucleus, which leads to the expression of \u003cem\u003eMHC I\u003c/em\u003e genes (fig. 3b). The \u003cem\u003eB2M\u003c/em\u003e gene binds to the \u003cem\u003eMHC class I\u003c/em\u003e (includes \u003cem\u003eHLA-A, B, C \u003c/em\u003ehaplotypes) (H3) and II (includes \u003cem\u003eHLA-DR, DQ\u003c/em\u003e, etc. haplotypes) (H4) genes in the nucleus and the complex is transferred to the cell surface to activate the immune system. So, \u003cem\u003eHLA\u003c/em\u003e Class II by presenting peptide antigens from the extracellular space and \u003cem\u003eHLA\u003c/em\u003e Class I by presenting peptide antigens from the cytoplasm to T-lymphocytes on the cell surface causes the activation of the immune system. The genes highlighted in yellow were observed in our results.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/24312df98287583db48509fa.png"},{"id":93013713,"identity":"ddcfcfe2-aeb3-4983-b5cd-c20786941f6f","added_by":"auto","created_at":"2025-10-08 07:28:28","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":294365,"visible":true,"origin":"","legend":"\u003cp\u003ePPI network between genes with high rank. This network shows the relationship between proteins encoded by genes expressed in Covid-19, MS, and AD. Pink node: AD gene; Blue node: MS gene; Green node: Covid-19 gene; Green and blue nodes: common genes between MS and Covid-19. The genes highlighted in pink and blue are suggested as blood biomarkers for early diagnosis of AD and MS, respectively.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/758386e337d930ef8e4c4926.png"},{"id":93011914,"identity":"1c1d2159-a745-45ad-8d44-5a1176b54eb9","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":397059,"visible":true,"origin":"","legend":"\u003cp\u003eMechanisms of SARS-CoV-2-Induced CNS Inflammation and Nerve Damage. (a) SARS-CoV-2 enters the CNS by crossing the blood-brain barrier through endocytosis, leading to microglia activation. Activated microglia release inflammatory cytokines, which prompt the activation of T and B cells, resulting in potential nerve damage (stages I-IV). (b) Damage to the blood-brain barrier causes inflammatory cytokines to enter the peripheral blood, increasing the expression of myosin genes (\u003cem\u003eMYO1F\u003c/em\u003e and \u003cem\u003eMYL6B\u003c/em\u003e), which play key roles in cell movement and immune responses (stages 1-8). SARS-CoV-2 also binds to Toll-like receptors (TLRs) in peripheral blood, triggering a cytokine storm and further nerve damage (stages a-i). This Fig. illustrates the progressive inflammation and immune response mechanisms that contribute to CNS damage and potential neurodegenerative, highlighting key biomarkers involved in Covid-19, MS, and AD.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/69e512c060d477fce0270cc1.png"},{"id":94254473,"identity":"20ceb4e1-00cf-479e-8ccb-8a3f0c9f8d92","added_by":"auto","created_at":"2025-10-24 07:32:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3091484,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/760315b3-3611-4776-8dbd-8346c4c81118.pdf"},{"id":93014850,"identity":"0cad0df7-5e59-440a-9fdb-25aca5f09d6d","added_by":"auto","created_at":"2025-10-08 07:44:27","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":80551,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese1.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/8832667d324e54eb002088a4.png"},{"id":93011899,"identity":"c250af2d-5281-44dd-8337-40591c48fdad","added_by":"auto","created_at":"2025-10-08 07:20:27","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":88905,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese2.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/03b9e14b92f8585148b45f25.png"},{"id":93011926,"identity":"37f52ab8-3b2d-4e90-8df0-efe787f35f16","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":87913,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese3.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/863587a0c0829cfdbd67798f.png"},{"id":93014098,"identity":"c2281608-a371-4904-ad98-6de4ff743e67","added_by":"auto","created_at":"2025-10-08 07:36:28","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":87706,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese4.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/a43a37385de1d8f4bf4f41fd.png"},{"id":93011927,"identity":"dee15662-830e-4963-aa4a-a6d582b32cf1","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":458326,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese5.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/44ae41cd03db406f450bf61e.png"},{"id":93013727,"identity":"63b07c3b-65ac-454c-ba49-8aa271467ba0","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":3438630,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese6.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/f5c4a6ab1d86ba1adec9ce20.png"},{"id":93011918,"identity":"1541a7e8-d757-4870-b7fb-4296373181e0","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":441951,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese7.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/67b712e570e266a1fa43f665.png"},{"id":93011916,"identity":"952ffddc-b99b-4e88-bcdd-cef6ad81086b","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":441951,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese8.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/e6e0fa5916f73fba746528d8.png"},{"id":93011913,"identity":"7bf4bc3e-1d73-4143-bf70-1bd9ac6dd6c6","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":15423,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurese9.png","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/8de0df302478403e8e7a6fe7.png"},{"id":93013722,"identity":"73a79786-38ca-42a9-a6be-b6e8fde10d10","added_by":"auto","created_at":"2025-10-08 07:28:28","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":6297286,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/219b2172635eba84f31e0a71.docx"},{"id":93013719,"identity":"fe5fd785-ae52-4173-92c7-fd225c3c49f8","added_by":"auto","created_at":"2025-10-08 07:28:28","extension":"xlsx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":27404,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/6faacfd2edb009c3f2e4aeb2.xlsx"},{"id":93011921,"identity":"a9a82003-ee00-4e23-ab9a-1d33e036dafe","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"xlsx","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":45641,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/ff1acba164dcac7cbe2ff965.xlsx"},{"id":93011924,"identity":"368c47fb-6428-49f4-b79e-2f0d37aeaf0d","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"xlsx","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":21384,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/547114bd7fcc898cb24a96df.xlsx"},{"id":93011936,"identity":"9362f781-d39a-4edb-8c1b-e4becfc85822","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"xlsx","order_by":14,"title":"","display":"","copyAsset":false,"role":"supplement","size":62659,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/793782dfc5de19017714cc02.xlsx"},{"id":93011932,"identity":"886de246-4c4f-4e46-96a7-a0fcac1b9103","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"xlsx","order_by":15,"title":"","display":"","copyAsset":false,"role":"supplement","size":115124,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/b24ab692c8e0fd09f9e03340.xlsx"},{"id":93011948,"identity":"c51b9df1-71b7-462d-9991-792362cb3702","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"xlsx","order_by":16,"title":"","display":"","copyAsset":false,"role":"supplement","size":46976,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/2688708638b300b362bff7d9.xlsx"},{"id":93011937,"identity":"0c2a4f92-fe17-4cff-84cb-849a8afaaedd","added_by":"auto","created_at":"2025-10-08 07:20:29","extension":"xlsx","order_by":17,"title":"","display":"","copyAsset":false,"role":"supplement","size":113118,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/d9a209848b56ebc5a6102cf1.xlsx"},{"id":93011935,"identity":"611f127b-658f-4ba5-ae72-a14fbcb3b3ce","added_by":"auto","created_at":"2025-10-08 07:20:28","extension":"xlsx","order_by":18,"title":"","display":"","copyAsset":false,"role":"supplement","size":203878,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/fca46a4b5689cc05fe2cb718.xlsx"},{"id":93013730,"identity":"280c5866-66d5-4eff-997b-4a93a3bc6079","added_by":"auto","created_at":"2025-10-08 07:28:29","extension":"xlsx","order_by":19,"title":"","display":"","copyAsset":false,"role":"supplement","size":17205,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7342199/v1/5840aa00e48cb1a3cb0d5d63.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Linking COVID-19 to Neurodegeneration: A Single-Cell Deep Learning Study of PBMCs in Multiple Sclerosis and Alzheimer’s Disease","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eSARS-CoV-2, the virus responsible for coronavirus disease 19 (Covid-19), has caused a global health crisis with widespread and persistent symptoms in affected individuals. Approximately 80% of Covid-19 patients experience long-term complications, including fatigue, anosmia, lung dysfunction, and various neurological disorders\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Among these symptoms, the neurological effects are particularly concerning due to their potential long-term impact. Emerging research indicates that Covid-19 not only induces immediate neurological symptoms but may also have lasting effects on the central nervous system (CNS), particularly in individuals with pre-existing neurological conditions or those at risk of developing them\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. For example, recent studies suggest a potential link between Covid-19 and MS, although the mechanisms are still unclear. Case reports describe new MS diagnosis or symptom onset following SARS-CoV-2 infection, such as a 47-year-old woman who developed symptoms three weeks post-infection\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Additionally, 51% of MS patients experienced symptom worsening, and 20% developed new symptoms during or after infection. Relapse rates also increased among relapsing-remitting MS patients during the Covid-19 \"at-risk\" period\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eImmune activation and inflammation within the CNS may be primary contributors to the neurological symptoms observed in SARS-CoV-2 patients. There are at least three possible routes through which SARS-CoV-2 can invade the brain:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eInfection in the nasal cavity could allow the virus to directly spread to the olfactory bulb via the olfactory nerves\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eThe virus may infect the eyes and subsequently reach the occipital cortex through the optic nerve\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eHigh viral titers in the respiratory tract could enable the virus to enter the bloodstream, spreading to multiple organs, including the CNS\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003cp\u003eReplication of the SARS-CoV-2 virus within host cells triggers an immune response by activating toll-like receptors (\u003cem\u003eTLRs\u003c/em\u003e) and \u003cem\u003eMDA5\u003c/em\u003e or \u003cem\u003eRIG-I\u003c/em\u003e, leading to the production of interferons (INF)\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-a). In response to interferon release, a group of immune cells accumulates in the peripheral blood, activating the \u003cem\u003eJAK-STAT1/2\u003c/em\u003e signaling pathway and contributing to the formation of a cytokine storm\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-c, d). This immune response can ultimately compromise the blood-brain barrier (BBB), allowing immune cells to infiltrate the CNS. Once in the CNS, these immune cells initiate severe immune responses, contributing to neurodegeneration and apoptosis, which can further stimulate multiple long-lasting immune responses within the CNS\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-e).\u003c/p\u003e\u003cp\u003eBefore entering the brain, viral pathogens must penetrate the blood-brain barrier (BBB)\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. SARS-CoV-2 may cross the BBB through several mechanisms, including the ACE2-mediated transcellular pathway, disruption of the BBB's tight junctions via the paracellular pathway, or intracellular transport, where viral particles are carried by infected cells, bypassing the BBB\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Once inside the CNS, the virus can trigger increased inflammatory responses, potentially leading to neurodegenerative diseases, particularly Multiple Sclerosis (MS).\u003c/p\u003e\u003cp\u003eMultiple sclerosis (MS) is a chronic inflammatory and demyelinating disease of the CNS, marked by immune cell infiltration, myelin destruction, and axonal damage, leading to neurological disability. Epstein-Barr virus infection is a key driver of MS, establishing a causal link between infection and neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. MS onset involves peripheral immune activation, followed by CNS invasion, supporting the \"outside-in\" theory. Mast cells disrupt the blood-brain barrier (BBB), allowing peripheral T cells, aided by myosin genes and MMPs, to infiltrate the CNS. Once inside, dendritic cells and CNS-resident cells release cytokines (TNF-α, IFN-γ, IL-17), which activate microglia and macrophages, causing myelin and axonal damage via neurotoxic molecules (MMPs, TNF-α, ROS, RNS) and CD8\u0026thinsp;+\u0026thinsp;T cell cytotoxicity. Astrocytes and NK cells, usually regulatory, also contribute to this process. As the BBB further weakens, more immune cells infiltrate, intensifying neural damage. This immune response accelerates MS progression and may explain how Covid-19 could worsen neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eSimilarly, Alzheimer\u0026rsquo;s disease (AD) is characterized by cognitive decline and neuronal loss, with genetic variants linked to immune response and neuroinflammation, such as \u003cem\u003eOAS1\u003c/em\u003e, increasing the risk for both AD and severe Covid-19\u003csup\u003e16\u003c/sup\u003e. Covid-19 can trigger cytokine storms, elevating pro-inflammatory cytokines like \u003cem\u003eIL-1β\u003c/em\u003e, \u003cem\u003eIL-6\u003c/em\u003e, and \u003cem\u003eTNF-α\u003c/em\u003e, which may contribute to synaptic dysfunction and neurodegeneration, particularly increasing the risk of AD in elderly individuals\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. The interplay between neuroinflammation and immune response, as seen in both AD and Covid-19, highlights the need for a deeper understanding of how these conditions influence brain health.\u003c/p\u003e\u003cp\u003eSingle-cell RNA sequencing (scRNA-seq) has significantly advanced our understanding of neurodegenerative disorders by revealing gene expression changes at the single-cell level, identifying key molecular alterations during disease progression\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Machine learning models help analyze these complex datasets, enabling precise biomarker identification\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. While traditional approaches rely on invasive procedures involving brain tissue or cerebrospinal fluid (CSF), they may miss subtle gene expression changes. Recently, there has been growing interest in using blood-based biomarkers for diseases like MS and AD. Blood samples offer a non-invasive alternative, promising earlier diagnosis, safer procedures, and improved disease management\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eRecent advances in blood-based biomarker discovery for AD provide a relevant framework for studying the neurological implications of COVID-19. In particular, the GeneDX-PBMC model, introduced by Talebi et al\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, demonstrated how single-cell RNA sequencing (scRNA-seq) data from peripheral blood mononuclear cells (PBMCs) can be combined with a deep learning framework\u0026mdash;integrating autoencoders, classifiers, and discriminators\u0026mdash;to identify both differentially expressed and subtle, non-differentially expressed genes associated with AD. This approach enabled the detection of systemic immune alterations linked to neuroinflammation, a mechanism that is also implicated in post-COVID neurological sequelae. Drawing on these methodological insights, our COVID-19 analysis adapts the same computational framework to investigate transcriptional signatures in PBMCs, aiming to uncover shared and disease-specific pathways connecting COVID-19 to neurodegenerative processes.\u003c/p\u003e\u003cp\u003eIn this paper, we leverage scRNA-seq profiling to explore the genetic links between Covid-19 and neurodegenerative diseases, specifically MS and AD. Traditional biomarker discovery methods, which rely on brain tissue, are not only invasive but also time-consuming and expensive. Furthermore, previous studies have primarily focused on differentially expressed genes (DEGs), often overlooking genes with subtle expression differences that may still play crucial roles in disease progression. To address these limitations, we analyzed scRNA-seq data from PBMCs of patients with MS, AD, and Covid-19. By employing a deep learning approach, we captured hidden nonlinear relationships between genes, effectively analyzed high-dimensional genomic data, and accurately ranked significant genes. Our model was applied separately to four cell types\u0026mdash;T-cells, B-cells, NK-cells, and monocytes\u0026mdash;identifying both DEGs and significant non-DEGs in each cell type. This provided deeper insights into the molecular mechanisms by which Covid-19 may contribute to these two neurodegenerative diseases.\u003c/p\u003e"},{"header":"2 Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Data collection\u003c/h2\u003e\u003cp\u003eThe publicly available scRNA-seq data for severe Covid-19, MS, and AD, with accession numbers GSE166992\u003csup\u003e23\u003c/sup\u003e, GSE138266\u003csup\u003e24\u003c/sup\u003e, and GSE181279\u003csup\u003e25\u003c/sup\u003e, respectively, were downloaded from the Gene Expression Omnibus (GEO) database \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Data for these conditions were collected from PBMCs using the same platform, ensuring comparability in the analysis. Additionally, a Covid-19 scRNA-seq dataset containing mild and asymptomatic cases, along with a severe flu dataset (under the GSE149689 accession number)\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e, was also included for comparative analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Data preprocessing\u003c/h2\u003e\u003cp\u003eFor quality control across all scRNA-seq datasets, genes with at least 200 features and non-zero counts, along with cells with a minimum count of three, were selected. Low-quality cells were removed from the dataset. After filtering, the data were normalized using a log-transformation to stabilize variance across the datasets. This preprocessing was performed using the Seurat package (version 5.1.1)\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e in R (version 3.6.1)\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. The results of the preprocessing steps are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSummary of Data Quality Control and Gene Statistics Across Different Conditions\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCondition\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of Samples\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNumber of Raw Cells\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNumber of Cells after QC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNumber of Cells After Doublet Removal\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eTotal Number of Genes\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCovid-19\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003e(Severe)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eD:5\u003c/p\u003e\u003cp\u003eN:3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e63895\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e37910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e31966\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e33538\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMultiple Sclerosis\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eD:5\u003c/p\u003e\u003cp\u003eN:3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e58300\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e49848\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e41979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e22164\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAlzheimer\u0026rsquo;s Disease\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eD:3\u003c/p\u003e\u003cp\u003eN:2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e36849\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e32177\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e27562\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e32738\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCovid-19\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003e(Mild \u0026amp; Asymptomatic)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eD:5\u003c/p\u003e\u003cp\u003eN:4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e53962\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e34607\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e29474\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e33538\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFlu (Severe)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eD:5\u003c/p\u003e\u003cp\u003eN:4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e39245\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e24697\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e21328\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e33538\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThis table presents the number of samples, raw cells, cells after quality control (QC), cells after doublet removal, and the total number of genes for each condition. The dataset includes samples from severe and mild cases of Covid-19, MS, AD, and severe flu. The conditions are categorized as Diseased (D) or Normal (N). QC and doublet removal steps were applied to ensure data quality before normalization using the Seurat package.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Method","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Overview:\u003c/h2\u003e\u003cp\u003eOur study employs a robust deep learning framework, directly adapted from the GeneDX-PBMC\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e methodology originally developed for Alzheimer\u0026rsquo;s disease biomarker discovery, to analyze PBMC scRNA-seq data and uncover gene expression patterns linked to neurodegenerative diseases. After data preparation, we clustered cell types into T cells, B cells, NK cells, and monocytes, identifying significant genes within each cluster (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-A). The framework integrates an autoencoder for feature extraction, a discriminator for adversarial training, and a classifier for disease prediction. The autoencoder reduces data dimensionality, the discriminator distinguishes between real and generated latent vectors, and the classifier predicts normal and diseased classes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B). While the architecture, training strategy, and gene-ranking approach remain identical to the original AD study, the input data and subsequent analyses are specific to the present COVID-19 investigation. We applied the model to publicly available PBMC scRNA-seq datasets from patients with COVID-19 (severe, mild, and asymptomatic), along with comparative datasets for multiple sclerosis (MS) and Alzheimer\u0026rsquo;s disease, to identify transcriptional patterns relevant to COVID-19 pathophysiology and its potential neurological impact.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Data Preparation\u003c/h2\u003e\u003cp\u003eTo rigorously evaluate our deep learning model, we employed two validation sets. Validation 1 involved splitting the dataset into 80% training and 20% testing, ensuring that the test set remained unseen during training. Validation 2 was performed by shuffling the dataset and excluding one diseased and one normal sample, which were used solely for testing model performance. This dual validation strategy allowed for comprehensive evaluation of the model's ability to generalize to unseen data.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Gene Selection and Cell-Type Clustering\u003c/h2\u003e\u003cp\u003eFollowing the setup of the training and validation sets, we shifted our focus to the training sets to identify significant genes critical for understanding disease progression. After removing batch effects using the IntegrateData function of the Seurat package, we employed K-means clustering to group similar cells, testing various values of k to determine the optimal number of clusters. The best fit was achieved with k\u0026thinsp;=\u0026thinsp;4, effectively capturing clustering patterns observed across different values of k. We consistently observed that similar normal and diseased cells were grouped within the same clusters.\u003c/p\u003e\u003cp\u003eTo further refine these clusters, we applied the SingleR package\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, which independently classified cells into four main types: T cells, B cells, NK cells, and monocytes, as depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-A. This classification aligned well with the clustering results from K-means. Within each cell type and disease, gene selection was performed using training set, applying an adjusted p-value threshold of \u0026lt;\u0026thinsp;0.05 and |log2FC| \u0026gt;0.01. This process included both differentially expressed genes (DEGs) and non-differentially expressed genes (non-DEGs).\u003c/p\u003e\u003cp\u003eSubsequently, a vector containing the expression levels of the significant genes was allocated to each cell. This vector was input into the neural network model, which was run independently for each cell type and disease, to identify specific patterns of gene expression.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Algorithm\u003c/h2\u003e\u003cp\u003eOur deep learning model integrates three distinct modules: an autoencoder, a classifier, and a discriminator, which collectively perform feature extraction, classification, and adversarial training using the PyTorch\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e library in Python In the feature extraction phase, the autoencoder compresses input data into a lower-dimensional latent space, focusing on deep feature extraction and dimensionality reduction, while the decoder reconstructs the original data from this compact representation. The classifier module then takes these latent representations to generate class predictions, incorporating a loss term that penalizes classification errors and enhances accuracy by prioritizing feature extraction over raw data preservation. Finally, the discriminator module is employed for adversarial training, learning to differentiate between real encoded representations and fake randomly generated latent vectors, thus ensuring the encoder produces meaningful features. This integrated approach facilitates precise classification, effective feature extraction, and robust adversarial training. Each component of the model is explained in detail below.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.5 First Module: Feature Extraction\u003c/h2\u003e\u003cp\u003eIn this module, the encoder and decoder work together to extract features and achieve dimensionality reduction, producing a more compact representation of the original data.\u003c/p\u003e\u003cp\u003eEncoder: The encoder compresses the input data into a lower-dimensional latent representation (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-a). This feature vector captures essential patterns of the original data while reducing its dimensionality. It consists of fully connected (linear) layers with batch normalization applied after each layer to enhance training stability and speed up convergence. Rectified Linear Unit (ReLU)\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e activation functions introduce non-linearity, and dropout is used as a regularization technique to prevent overfitting.\u003c/p\u003e\u003cp\u003eDecoder: The decoder reconstructs the input data from the latent representation (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-b). The objective during training is to minimize reconstruction loss, the difference between the original and reconstructed data. By reducing this loss, the autoencoder learns to focus on the most informative features. The decoder mirrors the encoder's structure, using fully connected layers with batch normalization and ReLU activation functions. Dropout is also applied to prevent overfitting. A final Sigmoid activation function ensures output values fall within the [0, 1] range. The Mean Absolute Error (MAE) loss module measures the absolute differences between the original and reconstructed data, guiding the training process by penalizing the model based on reconstruction errors.\u003c/p\u003e\u003cp\u003eMAE = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{1}{n}{\\sum\\:}_{i=1}^{n}\\left(yᵢ-ŷᵢ\\right)\\)\u003c/span\u003e\u003c/span\u003e (1)\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Second Module: Adversarial Training\u003c/h2\u003e\u003cp\u003e\u003cstrong\u003eClassifier\u003c/strong\u003e\u003cp\u003eThe classifier processes the encoded representation to generate class predictions (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-b). It consists of fully connected layers, with batch normalization applied to stabilize training and enhance convergence. Each fully connected layer is followed by a ReLU activation function to introduce non-linearity and capture complex patterns. To mitigate overfitting, a dropout layer is used after the initial layer. The final layer employs a LogSoftmax activation function to transform raw scores into a probability distribution suitable for multiclass classification tasks. The classifier uses cross-entropy loss to measure the difference between the predicted probability distribution and the true class distribution, guiding the model to minimize this difference during training.\u003c/p\u003e\u003c/p\u003e\u003cp\u003eCross Entropy Loss = -\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sum\\:}_{i=1}^{n}yᵢlog\\left(ŷᵢ\\right)\\)\u003c/span\u003e\u003c/span\u003e (2)\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.7 Third Module: Discriminator\u003c/h2\u003e\u003cp\u003eThe discriminator is utilized for adversarial training, a technique often employed in generative adversarial networks (GANs) and applied here in ADEP \u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Its primary objective is to differentiate between encoded representations produced by the encoder (real representations) and randomly generated latent vectors (fake representations) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-b). During training, the discriminator learns to distinguish between these real and fake representations.\u003c/p\u003e\u003cp\u003eThe discriminator's loss is typically calculated using binary cross-entropy or logistic loss, which measures the discrepancy between the discriminator\u0026rsquo;s predictions and the actual labels (real or fake). This loss functions as a quality control mechanism, ensuring that the encoder produces meaningful and informative representations of the input data.\u003c/p\u003e\u003cp\u003eThe discriminator consists of multiple fully connected (linear) layers, with batch normalization and ReLU activation functions applied after each layer to enhance training stability and capture complex patterns. The final layer uses a Sigmoid activation function to output a probability score within the range [0, 1], reflecting the likelihood that the input represents real encoded data. The Binary Cross-Entropy (BCE) loss module is employed to quantify the difference between the discriminator's predictions and the true labels. This loss guides the adversarial training process, helping the discriminator more effectively distinguish between real and generated representations. The BCE loss, also known as discriminator loss, is computed using the following formula:\u003c/p\u003e\u003cp\u003eBCE = - \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{1}{n}\\)\u003c/span\u003e\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sum\\:}_{i=1}^{n}\\left[yᵢ\\cdot\\:log\\left(ŷᵢ\\right)+\\left(1-yᵢ\\right)\\cdot\\:log\\left(1-ŷᵢ\\right)\\right]\\)\u003c/span\u003e\u003c/span\u003e (3)\u003c/p\u003e\u003cp\u003ewhere N is the number of samples in the dataset, y\u003csub\u003ei\u003c/sub\u003e​ is the ground truth label for sample i (1 for positive or real, 0 for negative or fake), and p\u003csub\u003ei\u003c/sub\u003e​ is the predicted probability that sample ii belongs to the positive class (real) as determined by the discriminator. This loss guides the adversarial training process, improving the discriminator\u0026rsquo;s ability to distinguish between real and generated representations.\u003c/p\u003e\u003cp\u003eLoss function: The core of this framework is the primary loss function, which integrates several distinct loss components. For our latent space, we include separate losses for fake and real representations, combined into an adversarial loss. Recognizing the inadequacy of adversarial loss alone, we supplement it with both decoder and encoder losses (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-c). The composite formulation of the total loss function is:\u003c/p\u003e\u003cp\u003eTotal Loss = (α \u0026middot; Autoencoder Loss) + (β \u0026middot; Classifier Loss) + (γ \u0026middot; Adversarial Loss)\u003c/p\u003e\u003cp\u003eHere, α, β, and γ are hyperparameters. This integration of loss terms facilitates precise classification and balances the critical tasks of discrimination, reconstruction, and feature extraction. This multifaceted approach enhances the overall performance and capabilities of our model.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.8 Hyperparameter Selection\u003c/h2\u003e\u003cp\u003eThe selection of hyperparameters, such as the learning rate, number of epochs, and dropout rates, was performed through a grid search approach combined with cross-validation. Specifically, we used a 5-fold cross-validation method on the training sets to tune the following key hyperparameters:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eLearning Rate: We experimented with learning rates in the range of 0.0001 to 0.001, ultimately selecting a value of 0.005, which provided the best balance between convergence speed and model performance without overfitting.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eEpochs: A total of 20 training epochs was chosen after evaluating the model\u0026rsquo;s performance across different values ranging from 10 to 30 epochs. We found that 20 epochs resulted in stable convergence without significant overfitting or underfitting.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eBatch Size: The optimal batch size was set to 32 after testing batch sizes of 16, 32, and 64, with 32 achieving the best trade-off between computational efficiency and model accuracy.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eDropout Rate: To prevent overfitting, we applied a dropout rate of 0.3 after experimenting with values between 0.1 and 0.3, with 0.1 minimizing validation loss while maintaining training stability.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThese hyperparameters were optimized independently for each cell type, with results showing that this configuration achieved robust performance across different cell types and disease conditions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.9 Gene Ranking\u003c/h2\u003e\u003cp\u003eAfter analyzing and classifying disease and normal samples using our deep learning model, we focused on ranking genes to identify the most significant ones. By multiplying the weight matrices of the neural network layers, we obtained a final weight matrix with two dimensions: columns representing disease or control conditions, and rows representing genes. For the disease column, we sorted this weight matrix in ascending order and ranked the genes accordingly (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-B-d). This method allowed us to prioritize genes associated with the disease, offering valuable insights for further biological research.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.10 Functional and Pathway Enrichment Analysis\u003c/h2\u003e\u003cp\u003eTo explore the biological significance of the identified genes the top-ranked genes from each cell type and condition were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis using the Gprofiler tools. This allowed us to categorize the genes into relevant biological processes, cellular components, and molecular functions. Additionally, we employed STRING to construct protein-protein interaction (PPI) networks, helping to identify key interactions among the candidate genes. This analysis revealed critical pathways involved in immune responses and neuroinflammatory processes. We also used GProfiler to assess the enrichment of specific pathways and to identify potential biomarkers linked to both Covid-19, MS, and AD. These findings are integrated into our gene ranking strategy and discussed in detail in the results section (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e-C).\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Results","content":"\u003cp\u003eIn this section, we present the outcomes of our comprehensive analysis of the genetic links between severe Covid-19, MS, and AD using scRNA-seq data. We detail the model training process, evaluation metrics, identification of significant genes, pathway analysis, candidate biomarker detection, and comparison of biomarker specificity across different Covid-19 variants and severe flu. The goal is to elucidate the molecular mechanisms potentially connecting these diseases and to identify unique biomarkers that could serve as specific indicators for early disease detection and progression monitoring.\u003c/p\u003e\u003cp\u003eIn our experiments, we configured the hyperparameters as follows: β and γ were both set to 1, while α was set to 0.5. These settings were carefully selected to balance the contributions of the autoencoder, classifier, and adversarial losses within the total loss function. Specifically, assigning equal weights to β and γ ensured that both the classifier and adversarial losses were prioritized equally, promoting robust classification performance and effective adversarial training. The value of α was set to 0.5 to moderately weigh the autoencoder loss, maintaining an optimal trade-off between feature extraction and reconstruction accuracy. This configuration allowed the model to efficiently learn the underlying patterns in the scRNA-seq data, while effectively distinguishing between disease and control samples. By balancing these loss components, our model achieved a stable training process, minimizing overfitting and enhancing generalizability across different cell types and disease conditions.\u003c/p\u003e\u003cp\u003eWhile the primary focus of our analysis centers on the genetic links between Covid-19 and MS, the investigation of Covid-19 and AD followed a similar approach. Both analyses utilized single-cell RNA sequencing (scRNA-seq) data and applied deep neural network models to identify relevant genetic interactions. Although we present the results for Covid-19 and AD in a more concise manner, the methodologies and analytical processes used for both disease comparisons were consistent, ensuring that the findings for each condition were derived from comparable techniques.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Model Training and Evaluation Across Cell Types\u003c/h2\u003e\u003cp\u003eThe performance of the model was evaluated using several metrics, including accuracy, AUROC (Area Under the Receiver Operating Characteristic curve), and AUPR (Area Under the Precision-Recall curve), for each cell type (T-cells, B-cells, NK-cells, and monocytes) across the two disease conditions: Covid-19, and MS. The model demonstrated high classification accuracy and significant improvement in identifying disease-associated gene expression patterns compared to traditional methods.\u003c/p\u003e\u003cp\u003eWe trained our deep learning model separately on each cell type\u0026mdash;T-cells, B-cells, NK-cells, and monocytes\u0026mdash;using data from Covid-19 and MS conditions. For each disease, 20% of the raw data was reserved for validation, while the remaining 80% was used for data processing and gene selection as the training set. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the number of cells and selected genes for each cell type across the different conditions. This distribution highlights the variability in cell counts and gene selection between conditions, which is crucial for understanding the model's performance on diverse datasets. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e also illustrates the performance measures of the deep learning method for each cell type and disease.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe number of cells and selected genes for each cell type under different conditions\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCondition\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eT-cell (cell/gene)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eB-cell\u003c/p\u003e\u003cp\u003e(cell/gene)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNK-cell\u003c/p\u003e\u003cp\u003e(cell/gene)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMonocyte\u003c/p\u003e\u003cp\u003e(cell/gene)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCovid-19\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e19153/251\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4166/359\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3854/310\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9694/432\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMultiple Sclerosis\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21139/1199\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4139/1112\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5227/1249\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10782/1410\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThis table provides a breakdown of cell counts and gene selection across T-cells, B-cells, NK-cells, and monocytes for Covid-19 and MS.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Identification of Significant Genes\u003c/h2\u003e\u003cp\u003eTo identify the most significant genes for each cell type and disease condition, we applied our gene ranking method. This method involves multiplying the weight matrices obtained from the neural network layers, resulting in a 2 \u0026times; n matrix, where 2 represents the disease and control conditions, and n is the number of significant genes identified for each cell type. Genes were ranked based on their weights in the disease column, with higher weights indicating greater relevance to the disease state. The ranked genes for each condition and cell type are presented in Supplementary Tables\u0026nbsp;1 through 12. These genes offer valuable insights into the potential molecular mechanisms underlying the diseases and highlight the genetic links between Covid-19, MS, AD, and neurodegenerative conditions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Biological Relevance and Pathway Analysis\u003c/h2\u003e\u003cp\u003eIntegrating gene ranking with pathway analysis allowed us to uncover the molecular mechanisms that potentially link Covid-19 and neurodegenerative diseases such as MS. By focusing on the top 100 ranked genes for each disease and cell type, we identified the most relevant genetic factors that may drive disease progression and interaction. This approach is particularly valuable for identifying subtle but crucial genetic interactions that might be overlooked in broader analyses.\u003c/p\u003e\u003cp\u003eSeveral of the top-ranked genes were found to be non-coding RNAs (such as \u003cem\u003eXIST, THUMPD3-AS1, LINC00513, MIF-AS1\u003c/em\u003e, etc.), which play significant regulatory roles by indirectly influencing the expression of genes involved in critical cellular processes such as proliferation, migration, apoptosis, and immune responses. Their inclusion underscores the complexity of genetic regulation in these diseases and the need for detailed functional exploration (Supplementary Table\u0026nbsp;13).\u003c/p\u003e\u003cp\u003eFollowing gene scoring, we selected the top 100 ranked genes for each disease and cell type and combined them, resulting in a set of 200 genes. The proteins encoded or regulated by these genes were analyzed using gProfiler\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e to identify enriched biological pathways, revealing several shared and distinct pathways across different cell types (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) (Supplementary Fig.\u0026nbsp;1\u0026ndash;4). In order to investigate more precisely, the Protein-Protein Interaction (PPI) network between the sets of genes in each cell type was drawn by STRING \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e (Supplementary Fig.\u0026nbsp;5\u0026ndash;8).\u003c/p\u003e\u003cp\u003eThese pathways reflect roles in immune regulation, response to infection, and inflammatory processes, providing deeper insights into how Covid-19 may influence the progression of neurodegenerative diseases, particularly through immune-related mechanisms.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePathway enrichment analysis identified for each cell type using GProfiler\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eB-cell\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003elog10 (p)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eT-cell\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eLog 10 (p)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNK-cell\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003elog10 (p)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eMonocyte\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003elog10 (p)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eInterferon\u003c/b\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\gamma\\:\\)\u003c/span\u003e\u003c/span\u003e\u003cb\u003esignaling\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-12.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eImmune response\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-7.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eImmune response\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-11.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePositive regulation of immune response\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-9.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eImmune response\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-8.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRegulation of reaction oxygen species metabolism\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-7.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eInterferon signaling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-8.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eRegulation of vesicle-mediated transport\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-7.25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eProgrammed cell death\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-7.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSARS-CoV2 signaling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-5.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eEpstein-barr virus infection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-7.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCytokine signaling in the immune system\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-6.12\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSARS-CoV2 infections\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-6.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCell motility\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-5.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eT-cell activation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-6.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eRegulation of receptor binding\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-5.45\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTNF\u003c/b\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e \u003cb\u003esignaling pathway\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-5.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePositive regulation of apoptosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-5.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSARS-CoV2 signaling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-5.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eInflammatory response to antigenic stimulus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-5.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMAPK signaling pathway\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-5.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTNF\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e signaling pathway\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-4.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eCellular response to cytokine stimuli\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-5.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eRegulation of protein kinase activity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e-4.95\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRegulation of INF signaling\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-5.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCytokine signaling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-4.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRegulation of B-cell activation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e-4.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePositive regulation of cell migration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e4.63\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThis table lists the key pathways associated with the top 200 genes (combined top 100 genes from Covid-19 and MS) for each cell type\u0026mdash;B-cells, T-cells, NK cells, and monocytes. \"log10 (p)\" is the p-value in log base 10.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Candidate Biomarker Detection\u003c/h2\u003e\u003cp\u003eTo identify the most significant genes for each disease and cell type, we refined our analysis by narrowing down the initial set of 100 genes to a smaller, more focused subset. This step is crucial for effective biomarker discovery, as working with a large number of genes can be impractical and may dilute the specificity of the findings. The top-ranked genes were sequentially evaluated using classical machine learning algorithms, such as support vector machines and random forests, to calculate their F1 scores. Genes were added incrementally until the combined F1 score of these selected genes closely approximated the performance of our deep-learning model. For example, in the case of Covid-19 monocyte, an F1 score of approximately 0.98 was achieved with just 5 genes, \u003cem\u003eXIST, HLA-DRB5, IGKV2-30, SULT1A1\u003c/em\u003e, and \u003cem\u003eRNF144B\u003c/em\u003e, nearly matching the deep learning model's F1 score of 0.989. This approach allowed us to efficiently pinpoint key genes while maintaining high predictive performance. The differential expression level of selected genes for Covid-19, MS, and AD in each cell type are illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. These genes represent potential biomarkers, showing a strong association with the respective diseases and highlighting their significance in disease mechanisms and progression. Among the selected genes, \u003cem\u003eXIST\u003c/em\u003e and \u003cem\u003eHLA-DRB5\u003c/em\u003e stand out for their presence in multiple cell types and their association with both diseases. \u003cem\u003eXIST\u003c/em\u003e has been identified as a hub lncRNA in MS \u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e and is implicated in immune response regulation, particularly in females, due to its role in activating interferon-α production via a \u003cem\u003eTLR7\u003c/em\u003e-dependent mechanism \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Elevated expression of \u003cem\u003eXIST\u003c/em\u003e (average log2FC of 12.79 in Covid-19 and 12.98 in MS) suggests its involvement in the immune response, particularly in the context of viral infections like Covid-19, where it may contribute to the observed female bias in disease severity \u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003cem\u003eHLA-DRB5\u003c/em\u003e, part of the major histocompatibility complex (MHC) class II\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, is crucial for presenting extracellular antigens to T-lymphocytes, thus enhancing immune responses\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Variations in \u003cem\u003eHLA-DRB5\u003c/em\u003e and other \u003cem\u003eHLA\u003c/em\u003e class II loci have been shown to influence the severity of Covid-19\u003csup\u003e42\u003c/sup\u003e and the development of secondary progressive MS\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Our findings show increased expression of \u003cem\u003eHLA-DRB5\u003c/em\u003e (avg_log2FC of 1.44 in Covid-19 and 2.10 in MS), indicating its potential role in both diseases and highlighting its strong association with HLA-DRB1 in the gene and PPI networks.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e4.5 Pathway and Protein-Protein Interaction Analysis\u003c/h2\u003e\u003cp\u003eTo further explore the functional implications of these candidate genes, we performed pathway analysis and constructed protein-protein interaction (PPI) networks using STRING. These analyses revealed key interactions and pathways that contribute to disease progression in both Covid-19 and MS.\u003c/p\u003e\u003cp\u003eFor example, in B cells, the \u003cem\u003eXIST (HNRNPU)\u003c/em\u003e gene can activate the \u003cem\u003eMAPK\u003c/em\u003e pathway through \u003cem\u003eDDX3X\u003c/em\u003e, leading to increased production of pro-inflammatory genes\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e illustrates this mechanism, where \u003cem\u003eXIST\u003c/em\u003e binds to \u003cem\u003eMYD88\u003c/em\u003e, activating several Toll-like receptors (\u003cem\u003eTLRs\u003c/em\u003e), including \u003cem\u003eTLR1, TLR2, TLR4, TLR5, and TLR6\u003c/em\u003e\u003csup\u003e45\u003c/sup\u003e. This activation triggers downstream pathways such as \u003cem\u003eMKK4\u003c/em\u003e and \u003cem\u003eIKKα/β\u003c/em\u003e, ultimately enhancing the activity of transcription factors like \u003cem\u003eJNK, FOS\u003c/em\u003e, and \u003cem\u003eJUN\u003c/em\u003e, which leads to elevated expression of pro-inflammatory genes. Additionally, \u003cem\u003eXIST\u003c/em\u003e activates endosomal \u003cem\u003eTLRs\u003c/em\u003e (\u003cem\u003eTLR3, TLR7, TLR8, and TLR9\u003c/em\u003e), resulting in the activation of the \u003cem\u003eNFκB\u003c/em\u003e pathway and subsequent release of inflammatory cytokines\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. This process can exacerbate immune responses and contribute to CNS inflammation. Furthermore, \u003cem\u003eDDX3X\u003c/em\u003e is recruited to viral replication sites, where it interacts with scaffold proteins of stress granules, including \u003cem\u003eG3BP1, G3BP2, IGF2BP2, IGF2BP3\u003c/em\u003e, and \u003cem\u003eYBX1\u003c/em\u003e\u003csup\u003e\u003cem\u003e47\u003c/em\u003e\u003c/sup\u003e. \u003cem\u003eDDX3X\u003c/em\u003e also participates in antiviral signaling pathways by regulating the activity of \u003cem\u003eTBK1, IKKε, TRAF3\u003c/em\u003e, and \u003cem\u003eIRF3\u003c/em\u003e, ultimately leading to the production of type I interferon and modulation of the immune response\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e(Supplementary Fig.\u0026nbsp;9).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eSimilarly, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e demonstrates the role of \u003cem\u003eHLA-DRB5\u003c/em\u003e in the immune response. It shows how interferon production activates the \u003cem\u003eJAK/STAT\u003c/em\u003e pathway, inducing \u003cem\u003eIRF-1\u003c/em\u003e, which then activates the \u003cem\u003eCIITA\u003c/em\u003e gene\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. This gene complex with other transcription factors (\u003cem\u003eATF1/CREB, NF-Y\u003c/em\u003e, and \u003cem\u003eRFX\u003c/em\u003e) promotes the expression of \u003cem\u003eMHC II\u003c/em\u003e genes\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. This mechanism enhances the immune system's ability to present antigens and activate T-lymphocytes, potentially contributing to the inflammatory processes observed in both diseases\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn NK cells, \u003cem\u003eMYL6B\u003c/em\u003e and \u003cem\u003eMYO1F\u003c/em\u003e, which are involved in cell motility. These genes facilitate the migration of immune cells from the blood to other tissues, such as the brain, where they can exacerbate inflammation and contribute to neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. The increased expression of \u003cem\u003eMYL6B\u003c/em\u003e and \u003cem\u003eMYO1F\u003c/em\u003e in Covid-19 and MS further supports their role in the progression of these diseases\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. These findings provide a comprehensive view of the molecular mechanisms linking Covid-19 and MS and underscore the potential of \u003cem\u003eXIST\u003c/em\u003e, \u003cem\u003eHLA-DRB5\u003c/em\u003e, and \u003cem\u003eMYO1F\u003c/em\u003e as candidate biomarkers for early disease detection. Further validation through in vitro and in vivo experiments is required to confirm their utility as blood-based markers for these diseases.\u003c/p\u003e\u003cp\u003eExtending our analysis to AD, we aimed to determine whether similar immune regulatory mechanisms might be influencing its pathogenesis. Although AD, Covid-19, and MS are distinct conditions, they share common pathways of immune dysregulation and inflammation, which are known to be modulated by sex-specific factors. AD is the most common form of dementia, comprising 60\u0026ndash;70% of all cases, and is increasingly recognized as a disorder involving both autoimmune and autoinflammatory mechanisms. In response to initial stimuli, amyloid-beta (Aβ) acts as an early reactive immunopeptide, triggering an innate immune cascade that leads to a chronic cycle of neuroinflammation and neuronal damage. Epidemiological data reveal a higher prevalence of AD in women, with a ratio of approximately 2:1 compared to men. This sex difference may be partly attributed to the age-related decline in estrogens post-menopause, along with incomplete X-chromosome inactivation that enhances the expression of immune-related genes in females.\u003c/p\u003e\u003cp\u003eTo explore potential overlaps between AD and the gene regulatory networks we identified in Covid-19 and MS, we analyzed an AD scRNA-seq dataset. While there was no direct overlap among the top-ranked genes, we identified key inflammatory genes in peripheral blood that are involved in pathways similar to those regulated by the Covid-19 and MS-associated genes. For instance, the genes \u003cem\u003eDDX3X\u003c/em\u003e and \u003cem\u003eXIST\u003c/em\u003e from Covid-19 were found to stimulate \u003cem\u003eYBX1\u003c/em\u003e in AD, while \u003cem\u003eHLA-C\u003c/em\u003e in AD interacts with \u003cem\u003eHLA-DQB1\u003c/em\u003e and \u003cem\u003eHLA-DRB5\u003c/em\u003e from Covid-19 and MS, respectively\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e,\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. These interactions are visualized in the STRING network (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), which highlights the shared molecular pathways and interactions between these diseases. This suggests that common mechanisms of immune activation and inflammation may underlie these conditions, contributing to their pathogenesis. This Fig. provides a consolidated view of our findings by displaying the interconnected pathways and genes identified across Covid-19, MS, and AD, as detailed in Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. This PPI network illustrates the roles of shared genes, such as \u003cem\u003eXIST\u003c/em\u003e, \u003cem\u003eDDX3X\u003c/em\u003e, and \u003cem\u003eHLA-DRB5\u003c/em\u003e, in immune regulation and inflammation, highlighting their potential as early biomarkers for neurodegenerative diseases. By mapping these genes across the conditions, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e encapsulates the immune activation mechanisms contributing to disease progression and underscores the need for further exploration into shared pathways of immune dysregulation.\u003c/p\u003e\u003cp\u003eOur findings indicate that genes such as \u003cem\u003eYBX1\u003c/em\u003e\u003csup\u003e56\u003c/sup\u003e, \u003cem\u003eDUSP1\u003c/em\u003e\u003csup\u003e57\u003c/sup\u003e, and \u003cem\u003eHLA-C\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e could serve as potential biomarkers for early detection of AD due to their role in inflammation-induced neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. Overexpression of \u003cem\u003eYBX1\u003c/em\u003e in T, B, and NK cells enhances immune cell survival\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e, while \u003cem\u003eDUSP1\u003c/em\u003e prevents apoptosis\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e, potentially protecting neuronal cells \u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e. Similarly, \u003cem\u003eHLA-C\u003c/em\u003e facilitates antigen presentation, which could exacerbate neuroinflammation when overexpressed\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. Given the significant roles these genes play in both systemic and CNS immune responses, further in vitro and in vivo validation is essential to confirm their potential as blood-based biomarkers for AD.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003e4.6 Comparison of Biomarker Specificity Across Covid-19 Variants and Severe Flu\u003c/h2\u003e\u003cp\u003eTo assess whether the candidate biomarkers identified for severe Covid-19 and MS are unique to these conditions, we applied our methodology to an additional scRNA-seq dataset comprising samples from mild and asymptomatic Covid-19 cases, as well as severe flu cases. Using the same approach of integrating gene ranking with pathway analysis, we aimed to uncover the molecular mechanisms potentially linking these conditions to neurodegenerative diseases like MS. These datasets were the only publicly available PBMCs resources suitable for such a comparison, ensuring the robustness and transparency of our analysis. Our findings revealed that there was no overlap between the significant genes identified in severe Covid-19 and MS with those detected in asymptomatic Covid-19 or severe flu cases. This suggests that the biomarkers associated with severe Covid-19 and MS are distinct and not shared with less severe forms of Covid-19 or with severe flu. Consequently, these genetic markers may serve as specific indicators for severe Covid-19 and MS, distinguishing them from both milder Covid-19 cases and other respiratory conditions. The detailed results of this comparative analysis are presented in Supplementary Tables\u0026nbsp;14 through 21.\u003c/p\u003e\u003c/div\u003e"},{"header":"5 Discussion","content":"\u003cp\u003eIn our study, we identified several key genes and pathways significantly associated with severe Covid-19 MS and AD. Notably, the differential expression of genes such as \u003cem\u003eXIST, HLA-DRB5, DDX3X, YBX1, DUSP1\u003c/em\u003e, and \u003cem\u003eHLA-C\u003c/em\u003e across various cell types suggests a potential link between these conditions and immune regulation. These findings highlight the role of non-coding RNAs and X-linked genes, indicating a complex interplay between genetic and environmental factors in disease progression. Given that \u003cem\u003eXIST\u003c/em\u003e and \u003cem\u003eDDX3X\u003c/em\u003e are involved in X-chromosome inactivation\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e and immune regulation\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e, their higher expression in females could enhance immune responses, potentially increasing susceptibility to autoimmune and neurodegenerative diseases like MS and AD while providing a protective effect against severe viral infections like Covid-19\u003csup\u003e66\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eSex has been shown to significantly influence immune responses at multiple levels\u0026mdash;chromosomal, epigenetic, and hormonal\u0026mdash;affecting disease onset, progression, and prognosis\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. Females generally exhibit stronger inflammatory, antiviral, and humoral immune responses compared to males due to variations in both innate and adaptive immunity\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e, as evidenced by their higher phagocytic activity of neutrophils and macrophages, and more competent antigen-presenting cells (APCs). Conversely, males have a greater number of natural killers cells\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e. These immunological disparities are reflected in epidemiological data showing that males experience higher severity and fatality rates from Covid-19, potentially due to higher expression levels of SARS-CoV-2 entry receptors like \u003cem\u003eACE2\u003c/em\u003e and \u003cem\u003eTMPRSS2\u003c/em\u003e.\u003c/p\u003e\u003cp\u003eTo further elucidate the interconnected mechanisms by which SARS-CoV-2 may influence CNS inflammation and immune responses in both Covid-19 and MS, we present Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. This Fig. synthesizes findings from our pathway analyses and highlights the molecular cascade initiated by SARS-CoV-2 that impacts the CNS. Specifically, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, SARS-CoV-2 infiltrates the CNS by crossing the blood-brain barrier, where it activates microglia\u0026mdash;the resident immune cells of the brain\u0026mdash;triggering the release of inflammatory cytokines and activating T and B cells, thereby contributing to nerve damage. On the other hand, the virus also increases the expression of cell motility-related genes, including \u003cem\u003eMYO1F\u003c/em\u003e and \u003cem\u003eMYL6B\u003c/em\u003e, facilitating the migration of immune cells from the blood into brain tissue, worsening neuroinflammation and promoting neurodegeneration in conditions such as Covid-19, MS and AD. These findings align with previous data on \u003cem\u003eXIST\u003c/em\u003e, \u003cem\u003eHLA-DRB5\u003c/em\u003e, and MYO1F, which were identified as potential biomarkers in the pathway analysis and are now underscored in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e as pivotal players in CNS immune responses and disease progression. Together, this evidence emphasizes the need for further study of these genes as blood-based markers, offering insight into potential early detection strategies for both diseases.\u003c/p\u003e\u003cp\u003eOverall, the sex-specific immune mechanisms we observed in Covid-19 and MS appear to extend to AD, underscoring the importance of personalized approaches in understanding and treating neurodegenerative diseases. Our study suggests that shared pathways of immune dysregulation and inflammation, influenced by sex and genetic factors, may underlie these conditions. Future research should aim to further elucidate these connections, potentially paving the way for targeted therapeutic interventions that take into account both genetic and sex-specific factors (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"6 Limitations of the study","content":"\u003cp\u003eThis study has some limitations that should be noted. We relied on publicly available scRNA-seq data, which may not fully represent all patient populations or disease conditions. The datasets were collected under different conditions, which might have introduced some variability, even though we addressed this during preprocessing. Additionally, since this is a computational study, experimental validation of the identified biomarkers is needed to confirm their biological relevance. Future research with larger and more diverse datasets will help strengthen and expand these findings.\u003c/p\u003e"},{"header":"7 Conclusion","content":"\u003cp\u003eThis study leveraged PBMC scRNA-seq to unravel the molecular connections between Covid-19, MS, and potentially AD. By analyzing gene expression at the single-cell level, we identified key genes such as \u003cem\u003eXIST\u003c/em\u003e, \u003cem\u003eDDX3X\u003c/em\u003e, and \u003cem\u003eHLA-DRB5\u003c/em\u003e, which are differentially expressed across various immune cell types in the context of Covid-19 and MS. These genes are involved in critical immune regulatory pathways, highlighting their role in modulating disease susceptibility and progression.\u003c/p\u003e\u003cp\u003eOur findings suggest a potential link between Covid-19 and the onset or exacerbation of MS and AD, as both diseases share common inflammatory pathways and mechanisms of immune dysregulation. This connection is particularly relevant for individuals with a family history of MS or AD, as the expression of the identified candidate genes in PBMCs could serve as early biomarkers for assessing the risk of developing these neurodegenerative diseases following Covid-19 infection.\u003c/p\u003e\u003cp\u003eImportantly, our analysis included both differentially expressed genes (DEGs) and non-DEGs, the latter proving significant even though they were often excluded in prior research. The candidate genes identified, including \u003cem\u003eXIST\u003c/em\u003e, \u003cem\u003eHLA-DRB5, MYO1F, YBX1, DUSP1\u003c/em\u003e, and \u003cem\u003eHLA-C\u003c/em\u003e, hold promise as biomarkers for the early detection of MS and AD. Evaluating their expression levels in blood samples could facilitate the identification of individuals at higher risk of progressing from Covid-19 to MS or developing AD, particularly in those with genetic predispositions. This approach offers a non-invasive and accessible means of early diagnosis, potentially enabling timely interventions that could mitigate disease progression.\u003c/p\u003e\u003cp\u003eIn summary, our study underscores the utility of PBMC scRNA-seq in uncovering the molecular links between Covid-19 and neurodegenerative diseases. The identified candidate genes lay the groundwork for developing blood-based diagnostic tools that can screen individuals with a family history of MS or AD, ultimately contributing to more personalized and preventive healthcare strategies. Future research should focus on validating these biomarkers in larger cohorts and exploring their clinical utility in early disease detection and risk assessment.\u003c/p\u003e"},{"header":"8 Theory and Calculation","content":"\u003cdiv id=\"Sec27\" class=\"Section2\"\u003e\u003ch2\u003e8.1 Theory\u003c/h2\u003e\u003cp\u003eThe theoretical foundation of this work is based on the intersection of infectious diseases and neurodegeneration, specifically the impact of severe Covid-19 on the immune system and its role in the progression of neurodegenerative diseases like MS and AD. Leveraging prior studies on immune dysregulation and inflammatory pathways, this work builds upon established knowledge of PBMC behavior and gene expression changes under pathological conditions. The study extends this theoretical framework by integrating novel insights into shared genetic and molecular mechanisms between Covid-19 and neurodegeneration, with a focus on identifying non-invasive biomarkers.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec28\" class=\"Section2\"\u003e\u003ch2\u003e8.2 Calculation\u003c/h2\u003e\u003cp\u003eThe practical implementation of the theory is achieved through a combination of advanced computational and statistical methods. These include:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eDeep Learning Models: Utilized autoencoders, classifiers, and adversarial training to process and analyze high-dimensional scRNA-seq data.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eDifferential Gene Expression Analysis: Identified DEGs and non-DEGs, emphasizing their functional roles in immune and neurodegenerative processes.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003ePathway Enrichment and Network Analysis: Conducted pathway enrichment to reveal significant biological pathways and constructed protein-protein interaction networks to identify key regulatory nodes.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eBiomarker Prioritization: Prioritized blood biomarkers over brain biomarkers for early\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"9 Glossary","content":"\u003cul type=\"disc\"\u003e\n \u003cli\u003ePBMC (Peripheral Blood Mononuclear Cells): A type of blood cell that includes lymphocytes (T cells, B cells, and NK cells) and monocytes, used for analyzing immune responses.\u003c/li\u003e\n \u003cli\u003escRNA-seq (Single-Cell RNA Sequencing): A technique to study gene expression at the individual cell level, allowing for detailed molecular profiling.\u003c/li\u003e\n \u003cli\u003eDEGs (Differentially Expressed Genes): Genes showing significant changes in expression levels between two conditions.\u003c/li\u003e\n \u003cli\u003eNon-DEGs: Genes that do not show significant changes in expression but may still play subtle, critical roles in disease processes.\u003c/li\u003e\n \u003cli\u003eAutoencoder: A type of artificial neural network used to learn efficient representations of data, particularly for dimensionality reduction.\u003c/li\u003e\n \u003cli\u003eAdversarial Training: A machine learning approach where models are trained to be robust against small perturbations or adversarial inputs.\u003c/li\u003e\n \u003cli\u003eBiomarker: A measurable indicator of a biological state or condition, used for diagnostics or monitoring disease progression.\u003c/li\u003e\n \u003cli\u003ePathway Enrichment Analysis: A computational method to identify biological pathways significantly impacted by a set of genes.\u003c/li\u003e\n \u003cli\u003eProtein-Protein Interaction (PPI) Network: A graphical representation of interactions between proteins, used to identify key functional hubs in biological systems.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll transcriptomic datasets analyzed in this study were obtained from the Gene Expression Omnibus (GEO) repository. The following accession numbers were used: GSE166992, GSE138266, GSE181279 and GSE149689. All data are publicly available at https://www.ncbi.nlm.nih.gov/geo/.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;Code Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe code for the deep learning model is accessible on GitHub repository [https://github.com/Shokoofeh3433/Deep-Learning-Model]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to thank the School of Biological Sciences, Institute for Research in Fundamental Sciences (IPM) and Computing Center of IPM in performing parallel computing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conducted without any external funding support.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors report no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interest regarding the research, authorship, and publication of this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConceptualization:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData curation:\u003c/strong\u003e Shokoofeh Ghiam.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFormal analysis:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor, Changiz Eslahchi.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInvestigation:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh, Changiz Eslahchi.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethodology:\u003c/strong\u003e Shokoofeh Ghiam, Changiz Eslahchi.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProject administration:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResources:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSoftware:\u003c/strong\u003e Asiyeh Mirzaei, Mohammad Shirinpoor.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupervision:\u003c/strong\u003e Shokoofeh Ghiam, Changiz Eslahchi.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eValidation:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Pourya Naderi-Yeganeh, Mohammad Shirinpoor.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWriting \u0026ndash; original draft:\u003c/strong\u003e Asiyeh Mirzaei, Shokoofeh Ghiam, Mohammad Shirinpoor.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWriting \u0026ndash; review \u0026amp; editing:\u003c/strong\u003e Pourya Naderi-Yeganeh, Shokoofeh Ghiam, Asiyeh Mirzaei, Mohammad Shirinpoor, Changiz Eslahchi.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eDhama K, Patel SK, Pathak M, et al. An update on SARS-CoV-2/COVID-19 with particular reference to its clinical pathology, pathogenesis, immunopathology and mitigation strategies. Travel Medicine and Infectious Disease. 2020; 37:101755. doi: 10.1016/j.tmaid.2020.101755\u003c/li\u003e\n\u003cli\u003eCamerlingo C. COVID-19 and neurodegenerative diseases. European Review. June 28, 2022. http://www.europeanreview.org/article/29093\u003c/li\u003e\n\u003cli\u003eSpudich S, Nath A. Nervous system consequences of COVID-19. Science. 2022;375(6578):267-269. doi:10.1126/science. abm2052\u003c/li\u003e\n\u003cli\u003eSarwar S, Rogers S, Mohamed AS, et al. Multiple Sclerosis Following SARS-CoV-2 Infection: A Case Report and Literature Review. Cureus. Published online October 25, 2021. doi:10.7759/cureus.19036\u003c/li\u003e\n\u003cli\u003eMacDougall M, El-Hajj Sleiman J, Beauchemin P, Rangachari M. SARS-CoV-2 and Multiple Sclerosis: Potential for Disease Exacerbation. Frontiers in Immunology. 2022;13. doi:10.3389/fimmu.2022.871276\u003c/li\u003e\n\u003cli\u003eBassetti M, Vena A, Giacobbe DR. The novel Chinese coronavirus (2019‐nCoV) infections: Challenges for fighting the storm. European Journal of Clinical Investigation. 2020;50(3). doi:10.1111/eci.13209\u003c/li\u003e\n\u003cli\u003eAhmed SF, Quadeer AA, McKay MR. Preliminary Identification of Potential Vaccine Targets for the COVID-19 Coronavirus (SARS-CoV-2) Based on SARS-CoV Immunological Studies. Viruses. 2020;12(3):254. doi:10.3390/v12030254\u003c/li\u003e\n\u003cli\u003eHoang HD, Naeli P, Alain T, Jafarnejad SM. Mechanisms of impairment of interferon production by SARS-CoV-2. Biochemical Society Transactions. 2023;51(3):1047-1056. doi:10.1042/bst20221037\u003c/li\u003e\n\u003cli\u003eYang L, Xie X, Tu Z, Fu J, Xu D, Zhou Y. The signal pathways and treatment of cytokine storm in COVID-19. Signal Transduction and Targeted Therapy. 2021;6(1). doi:10.1038/s41392-021-00679-0\u003c/li\u003e\n\u003cli\u003eZang X, Chen S, Zhu J, Ma J, Zhai Y. The Emerging Role of Central and Peripheral Immune Systems in Neurodegenerative Diseases. Frontiers in Aging Neuroscience. 2022;14. doi:10.3389/fnagi.2022.872134\u003c/li\u003e\n\u003cli\u003eHui DS, I Azhar E, Madani TA, et al. The continuing 2019-nCoV epidemic threat of novel coronaviruses to global health \u0026mdash; The latest 2019 novel coronavirus outbreak in Wuhan, China. International Journal of Infectious Diseases. 2020; 91:264-266. doi: 10.1016/j.ijid.2020.01.009\u003c/li\u003e\n\u003cli\u003eAhmad T, Khan M, Haroon, et al. COVID-19: Zoonotic aspects. Travel Medicine and Infectious Disease. 2020; 36:101607. doi: 10.1016/j.tmaid.2020.101607\u003c/li\u003e\n\u003cli\u003eBjornevik K, M\u0026uuml;nz C, Cohen JI, Ascherio A. Epstein\u0026ndash;Barr virus as a leading cause of multiple sclerosis: mechanisms and implications. Nature Reviews Neurology. Published online February 9, 2023. doi:10.1038/s41582-023-00775-5\u003c/li\u003e\n\u003cli\u003eBalasa R, Barcutean L, Mosora O, Manu D. Reviewing the Significance of Blood\u0026ndash;Brain Barrier Disruption in Multiple Sclerosis Pathology and Treatment. International Journal of Molecular Sciences. 2021;22(16):8370. doi:10.3390/ijms22168370\u003c/li\u003e\n\u003cli\u003ePark HJ, Jung H. Neuro-immune interactions at single-cell resolution in neurodevelopmental, infectious, and neurodegenerative diseases. Animal Cells and Systems. 2022;26(4):137-147. doi:10.1080/19768354.2022.2110937\u003c/li\u003e\n\u003cli\u003eKumar D, Jahan S, Khan A, et al. Neurological Manifestation of SARS-CoV-2 Induced Inflammation and Possible Therapeutic Strategies Against COVID-19. Molecular Neurobiology. Published online March 14, 2021. doi:10.1007/s12035-021-02318-9\u003c/li\u003e\n\u003cli\u003eLi X, Wang CY. From bulk, single-cell to spatial RNA sequencing. International Journal of Oral Science. 2021;13(1). doi:10.1038/s41368-021-00146-0\u003c/li\u003e\n\u003cli\u003eAl-Tashi Q, Saad MB, Muneer A, et al. Machine Learning Models for the Identification of Prognostic and Predictive Cancer Biomarkers: A Systematic Review. International Journal of Molecular Sciences. 2023;24(9):7781. doi:10.3390/ijms24097781\u003c/li\u003e\n\u003cli\u003eBiernacki T, Kokas Z, Sandi D, et al. Emerging Biomarkers of Multiple Sclerosis in the Blood and the CSF: A Focus on Neurofilaments and Therapeutic Considerations. International Journal of Molecular Sciences. 2022;23(6):3383. doi:10.3390/ijms23063383\u003c/li\u003e\n\u003cli\u003eJ. Huang, M. Khademi, L. Fugger, \u0026Ouml;. Lindhe, L. Novakova, M. Axelsson, C. Malmestr\u0026ouml;m, C. Constantinescu, J. Lycke, F. Piehl, T. Olsson, I. Kockum, Inflammation-related plasma and CSF biomarkers for multiple sclerosis, Proc. Natl. Acad. Sci. U.S.A.117 (23) 12952-12960, https://doi.org/10.1073/pnas.1912839117 (2020).\u003c/li\u003e\n\u003cli\u003eThompson EA, Cascino K, Ordonez AA, et al. Metabolic programs define dysfunctional immune responses in severe COVID-19 patients. Cell Reports. 2021;34(11):108863. doi: 10.1016/j.celrep.2021.108863\u003c/li\u003e\n\u003cli\u003eTalebi H, Ghiam S, Mirzaei Koli A, Naderi Yeganeh P, Eslahchi C. GeneDX-PBMC: An adversarial autoencoder framework for unlocking Alzheimer\u0026rsquo;s disease biomarkers using blood single-cell RNA sequencing data. \u003cem\u003eComput Biol Med\u003c/em\u003e. 2025; 192:110283. doi: 10.1016/j.compbiomed.2025.110283\u003c/li\u003e\n\u003cli\u003eSchafflick D, Xu CA, Hartlehnert M, et al. Integrated single cell analysis of blood and cerebrospinal fluid leukocytes in multiple sclerosis. Nature Communications. 2020;11(1). doi:10.1038/s41467-019-14118-w\u003c/li\u003e\n\u003cli\u003eXu H, Jia J. Single-Cell RNA Sequencing of Peripheral Blood Reveals Immune Cell Signatures in Alzheimer\u0026rsquo;s Disease. Frontiers in Immunology. 2021;12. doi:10.3389/fimmu.2021.645666\u003c/li\u003e\n\u003cli\u003eBarrett T, Wilhite SE, Ledoux P, et al. NCBI GEO: archive for functional genomics data sets\u0026mdash;update. Nucleic Acids Research. 2012;41(D1): D991-D995. doi:10.1093/nar/gks1193\u003c/li\u003e\n\u003cli\u003eLee JS, Park S, Jeong HW, et al. Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19. Science Immunology. 2020;5(49). doi:10.1126/sciimmunol. abd1554\u003c/li\u003e\n\u003cli\u003eHao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573-3587.e29. doi: 10.1016/j.cell.2021.04.048\u003c/li\u003e\n\u003cli\u003eHao et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nature Biotechnology (2023) [Seurat V5]\u003c/li\u003e\n\u003cli\u003eR: A Language and Environment for Statistical Computing: Reference Index.; 2010.\u003c/li\u003e\n\u003cli\u003eAran D, Looney AP, Liu L, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nature Immunology. 2019;20(2):163-172. doi:10.1038/s41590-018-0276-y\u003c/li\u003e\n\u003cli\u003eChintala AP Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, Soumith. PyTorch: An imperative style, high-performance deep learning library. NeurIPS. 2019;32(1).\u003c/li\u003e\n\u003cli\u003eCreswell A, White T, Dumoulin V, Arulkumaran K, Sengupta B, Bharath AA. Generative Adversarial Networks: An Overview. IEEE Signal Processing Magazine. 2018;35(1):53-65. doi:10.1109/msp.2017.2765202\u003c/li\u003e\n\u003cli\u003eSzklarczyk D, Kirsch R, Koutrouli M, et al. The STRING database in 2023: protein\u0026ndash;protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Research. 2022;51(D1): D638-D646. doi:10.1093/nar/gkac1000\u003c/li\u003e\n\u003cli\u003eKolberg L, Raudvere U, Kuzmin I, Adler P, Vilo J, Peterson H. g: Profiler\u0026mdash;interoperable web service for functional enrichment analysis and gene identifier mapping (2023 update). Nucleic Acids Research. 2023;51(W1): W207-W212. doi:10.1093/nar/gkad347\u003c/li\u003e\n\u003cli\u003eDing Y, Li T, Yan X, et al. Identification of hub lncRNA ceRNAs in multiple sclerosis based on ceRNA mechanisms. Molecular Genetics and Genomics. 2021;296(2):423-435. doi:10.1007/s00438-020-01750-1\u003c/li\u003e\n\u003cli\u003eSouyris M, Cenac C, Azar P, et al. TLR7 escapes X chromosome inactivation in immune cells. Science Immunology. 2018;3(19). doi:10.1126/sciimmunol. aap8855\u003c/li\u003e\n\u003cli\u003eYu B, Qi Y, Li R, Shi Q, Satpathy AT, Chang HY. B cell-specific XIST complex enforces X-inactivation and restrains atypical B cells. Cell. 2021;184(7):1790-1803.e17. doi: 10.1016/j.cell.2021.02.015\u003c/li\u003e\n\u003cli\u003eGumkowska-Sroka, O., Kotyla, K., \u0026amp; Kotyla, P. (2024). Immunogenetics of Systemic Sclerosis. Genes, 15(5), 586. https://doi.org/10.3390/genes15050586\u003c/li\u003e\n\u003cli\u003eMedhasi S, Chantratita N. Human Leukocyte Antigen (HLA) System: Genetics and Association with Bacterial and Viral Infections. Journal of Immunology Research. 2022; 2022:1-15. doi:10.1155/2022/9710376\u003c/li\u003e\n\u003cli\u003eHoseinnezhad T, Soltani N, Ziarati S, Behboudi E, Mousavi MJ. The role of HLA genetic variants in COVID‐19 susceptibility, severity, and mortality: A global review. Journal of Clinical Laboratory Analysis. 2024;38(1-2). doi:10.1002/jcla.25005\u003c/li\u003e\n\u003cli\u003ePrapas P, Anagnostouli M. Macrophages and HLA-Class II Alleles in Multiple Sclerosis: Insights in Therapeutic Dynamics. International Journal of Molecular Sciences. 2024;25(13). doi:10.3390/ijms25137354\u003c/li\u003e\n\u003cli\u003eKovarik P, Bestehorn A and Fesselet J (2021) Conceptual Advances in Control of Inflammation by the RNA-Binding Protein Tristetraprolin. \u003cem\u003eFront. Immunol.\u003c/em\u003e 12:751313. doi: 10.3389/fimmu.2021.751313\u003c/li\u003e\n\u003cli\u003eAsami J, Shimizu T. Structural and functional understanding of the toll‐like receptors. Protein Science. 2021;30(4):761-772. doi:10.1002/pro.4043\u003c/li\u003e\n\u003cli\u003eBost C, Arleevskaya MI, Brooks WH, Plaza S, Guery JC, Renaudineau Y. Long non-coding RNA Xist contribution in systemic lupus erythematosus and rheumatoid arthritis. Clinical Immunology. 2022; 236:108937. doi: 10.1016/j.clim.2022.108937\u003c/li\u003e\n\u003cli\u003eCiccosanti F, Di Rienzo M, Romagnoli A, et al. Proteomic analysis identifies the RNA helicase DDX3X as a host target against SARS-CoV-2 infection. Antiviral Research. 2021; 190:105064. doi: 10.1016/j.antiviral.2021.105064\u003c/li\u003e\n\u003cli\u003eHern\u0026aacute;ndez-D\u0026iacute;az T, Valiente-Echeverr\u0026iacute;a F, Soto-Rifo R. RNA Helicase DDX3: A Double-Edged Sword for Viral Replication and Immune Signaling. Microorganisms. 2021;9(6):1206. doi:10.3390/microorganisms9061206\u003c/li\u003e\n\u003cli\u003eTong J, Zhang W, Chen Y, Yuan Q, Qin N-N and Qu G (2022) The Emerging Role of RNA Modifications in the Regulation of Antiviral Innate Immunity. \u003cem\u003eFront. Microbiol\u003c/em\u003e. 13:845625. doi: 10.3389/fmicb.2022.845625\u003c/li\u003e\n\u003cli\u003eMays Mohammed Salih, Chi G Weindel, Eric Malekos, Lisa Sudek, Sol Katzman, Cory J Mabry, Aja K Coleman, Sikandar Azam, Robert Watson, Kristin Patrick, Susan Carpenter, bioRxiv 2023.10.12.562050; doi: https://doi.org/10.1101/2023.10.12.562050\u003c/li\u003e\n\u003cli\u003eRenata Mezyk-Kopec, Lambert Potin, J. Emiliano Gomez Medellin, Calixto M. Salles, Melody A. Swartz; TGF-\u0026beta; Signaling Prevents MHC Class II\u0026ndash;Expressing Lymphatic Endothelial Cells from Reactivating Human Allogenic Memory CD4+ T Cells. J Immunol 1 September 2023; 211 (5): 782\u0026ndash;790. https://doi.org/10.4049/jimmunol.2200216\u003c/li\u003e\n\u003cli\u003eMatozo T, Kogachi L, de Alencar BC. Myosin motors on the pathway of viral infections. Cytoskeleton. 2022;79(6-8):41-63. doi:10.1002/cm.21718\u003c/li\u003e\n\u003cli\u003eMatozo T, Kogachi L, de Alencar BC. Myosin motors on the pathway of viral infections. Cytoskeleton. 2022;79(6-8):41-63. doi:10.1002/cm.21718\u003c/li\u003e\n\u003cli\u003eGhiam S, Eslahchi C, Shahpasand K, Habibi-Rezaei M, Gharaghani S. Exploring the role of non-coding RNAs as potential candidate biomarkers in the cross-talk between diabetes mellitus and Alzheimer\u0026rsquo;s disease. Frontiers in Aging Neuroscience. 2022;14. doi:10.3389/fnagi.2022.955461\u003c/li\u003e\n\u003cli\u003eGhiam S, Eslahchi C, Shahpasand K, Habibi-Rezaei M, Gharaghani S. Identification of repurposed drugs targeting significant long non-coding RNAs in the cross-talk between diabetes mellitus and Alzheimer\u0026rsquo;s disease. Scientific Reports. 2022;12(1). doi:10.1038/s41598-022-22822-9\u003c/li\u003e\n\u003cli\u003eKloetgen A, Duggimpudi S, Schuschel K, et al. YBX1 Indirectly Targets Heterochromatin-Repressed Inflammatory Response-Related Apoptosis Genes through Regulating CBX5 mRNA. International Journal of Molecular Sciences. 2020;21(12). doi:10.3390/ijms21124453\u003c/li\u003e\n\u003cli\u003ePan AL, Audrain M, Sakakibara E, et al. Dual-Specificity Protein Phosphatase 4 (DUSP4) Overexpression Improves Learning Behavior Selectively in Female 5xFAD Mice, and Reduces \u0026beta;-Amyloid Load in Males and Females. Cells. 2022;11(23). doi:10.3390/cells11233880\u003c/li\u003e\n\u003cli\u003eCătană CS, Marta MM, Văleanu M, Dican L, Crișan CA. Human Leukocyte Antigen and microRNAs as Key Orchestrators of Mild Cognitive Impairment and Alzheimer\u0026rsquo;s Disease: A Systematic Review. International Journal of Molecular Sciences. 2024;25(15). doi:10.3390/ijms25158544\u003c/li\u003e\n\u003cli\u003eHayden AN, Brandel KL, Merlau PR, et al. Behavioral Screening of Conserved RNA-Binding Proteins Reveals CEY-1/YBX RNA-Binding Protein Dysfunction Leads to Impairments in Memory and Cognition. Cold Spring Harbor Laboratory; 2024. Accessed October 15, 2024. http://dx.doi.org/10.1101/2024.01.05.574402\u003c/li\u003e\n\u003cli\u003eShao LL, Gao MM, Gong JX, Yang LY. DUSP1 regulates hippocampal damage in epilepsy rats via ERK1/2 pathway. Journal of Chemical Neuroanatomy. 2021; 118:102032. doi: 10.1016/j.jchemneu.2021.102032\u003c/li\u003e\n\u003cli\u003eYang J, Sun L, Han J, Zheng W, Peng W. DUSP1/MKP-1 regulates proliferation and apoptosis in keratinocytes through the ERK/Elk-1/Egr-1 signaling pathway. Life Sciences. 2019; 223:47-53. doi: 10.1016/j.lfs.2019.03.018\u003c/li\u003e\n\u003cli\u003eP\u0026eacute;rez-Sen R, Queipo MJ, Gil-Redondo JC, et al. Dual-Specificity Phosphatase Regulation in Neurons and Glial Cells. International Journal of Molecular Sciences. 2019;20(8):1999. doi:10.3390/ijms20081999\u003c/li\u003e\n\u003cli\u003eGross CC, Meyer C, Bhatia U, et al. CD8+ T cell-mediated endotheliopathy is a targetable mechanism of neuro-inflammation in Susac syndrome. Nature Communications. 2019;10(1). doi:10.1038/s41467-019-13593-5\u003c/li\u003e\n\u003cli\u003eFang H, Disteche CM, Berletch JB. X Inactivation and Escape: Epigenetic and Structural Features. Frontiers in Cell and Developmental Biology. 2019;7. doi:10.3389/fcell.2019.00219\u003c/li\u003e\n\u003cli\u003eRyan CS and Schr\u0026ouml;der M (2022) The human DEAD-box helicase DDX3X as a regulator of mRNA translation. Front. Cell Dev. Biol. 10:1033684. doi: 10.3389/fcell.2022.1033684\u003c/li\u003e\n\u003cli\u003eChlamydas S, Papavassiliou AG, Piperi C. Epigenetic mechanisms regulating COVID-19 infection. Epigenetics. 2020;16(3):263-270. doi:10.1080/15592294.2020.1796896\u003c/li\u003e\n\u003cli\u003eDodd KC and Menon M (2022) Sex bias in lymphocytes: Implications for autoimmune diseases. \u003cem\u003eFront. Immunol.\u003c/em\u003e 13:945762. doi: 10.3389/fimmu.2022.945762\u003c/li\u003e\n\u003cli\u003eTakahashi T, Iwasaki A. Sex differences in immune responses. \u003cem\u003eScience\u003c/em\u003e. 2021;371(6527):347-348. doi:10.1126/science. abe7199\u003c/li\u003e\n\u003cli\u003eJacobsen H and Klein SL (2021) Sex Differences in Immunity to Viral Infections. \u003cem\u003eFront. Immunol.\u003c/em\u003e 12:720952. doi: 10.3389/fimmu.2021.720952\u003c/li\u003e\n\u003cli\u003eKucuksezer UC, Aktas Cetin E, Esen F, Tahrali I, Akdeniz N, Gelmez MY, Deniz G. The Role of Natural Killer Cells in Autoimmune Diseases. \u003cem\u003eFrontiers in Immunology\u003c/em\u003e. 2021; 12:622306. doi:10.3389/fimmu.2021.622306\u003c/li\u003e\n\u003cli\u003eForsblom E, Sil\u0026eacute;n S, Kortela E, et al. Male predominance in disease severity and mortality in a low Covid-19 epidemic and low case-fatality area \u0026ndash; a population-based registry study. Infectious Diseases. 2021;53(10):789-799. doi:10.1080/23744235.2021.1936157\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7342199/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7342199/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCOVID-19 has increasingly been linked to neurological complications that may overlap with those observed in neurodegenerative and autoimmune diseases.\u003cbr\u003e\nIn this study, we analyzed single-cell RNA-sequencing data from peripheral blood mononuclear cells (PBMCs) of patients with COVID-19, multiple sclerosis (MS), and Alzheimer’s disease (AD).\u003cbr\u003e\nUsing a deep neural network combining autoencoders and adversarial learning, we uncovered distinct and shared transcriptional signatures across these conditions.\u003cbr\u003e\n Top-ranked genes—including \u003cem\u003eHLA-DRB5\u003c/em\u003e, \u003cem\u003eXIST\u003c/em\u003e, and \u003cem\u003eDDX3X\u003c/em\u003e—were not necessarily differentially expressed but demonstrated strong functional relevance through pathway enrichment and protein interaction analysis, highlighting latent biomarkers often missed by traditional DEG-based methods.\u003cbr\u003e\n Importantly, these candidate genes may aid in the detection of MS and AD among individuals with severe COVID-19 and a family history of these disorders, offering a non-invasive strategy for risk stratification and early intervention.\u003cbr\u003e\nOur findings underscore the value of PBMC-based scRNA-seq and deep neural network frameworks for discovering non-invasive biomarkers and highlight systemic and neuroinflammatory pathways that may connect COVID-19 to long-term neurological outcomes.\u003cbr\u003e\n This integrative approach may pave the way for novel diagnostic and therapeutic strategies, emphasizing the shared immunological underpinnings of these complex diseases.\u003c/p\u003e","manuscriptTitle":"Linking COVID-19 to Neurodegeneration: A Single-Cell Deep Learning Study of PBMCs in Multiple Sclerosis and Alzheimer’s Disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-08 07:20:22","doi":"10.21203/rs.3.rs-7342199/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d2dd3dd2-bee9-477f-8007-1bb9d3d3fdfb","owner":[],"postedDate":"October 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":55770080,"name":"Health sciences/Biomarkers"},{"id":55770081,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":55770082,"name":"Health sciences/Neurology"},{"id":55770083,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2025-10-24T07:24:04+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-08 07:20:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7342199","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7342199","identity":"rs-7342199","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}