Modeling Spatiotemporal Immune Dynamics in Inflammatory Diseases Using Integrative Multi-Modal Analytics and High-Dimensional Longitudinal Datasets

*Corresponding author: Adetayo Folasole. Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0. Modeling Spatiotemporal Immune Dynamics in Inflammatory Diseases Using Integrative Multi-Modal Analytics and High-Dimensional Longitudinal Datasets Adetayo Folasole 1, *, Patience Emanre Eshua 1 and Oluwagbemisola Elizabeth Elesho 2 1 Department of Computing, East Tennessee State University, United States of America. 2 Department of Biology, Georgia State University, United States of America. World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 Publication history: Received on 05 July 2025; revised on 12 August 2025; accepted on 14 August 2025 Article DOI: https://doi.org/10.30574/wjbphs.2025.23.2.0754 Abstract Comprehending the dynamic response of the immune system in inflammatory diseases is a major task due to the dynamic complexity, the cellular heterogeneity, and responses depend on the spatial context. The classic analytical methods are insufficient to handle the complex dynamics of the immune system, and especially of those diseases that advance, relapse, or remit temporally and spatially. To fill this void, the development of multi-modal analytics combined with high-dimensional, longitudinal data establishes a transformative framework for deciphering spatiotemporal immune signatures and regulatory networks. The aim of this study was to develop a global model system that uses integrative multi-omics data, imaging, clinical information, and spatial transcriptomics to model the immune dynamics in chronic inflammation including, e.g., rheumatoid arthritis (RA), IBD and SLE. By leveraging a number of sophisticated statistical learning algorithms, such as manifold alignment, tensor decomposition, and dynamic Bayesian networks, the model captures the sequential progression of disease progression, rewiring of immune cells, and microenvironmental cues over time. A key innovation is the method’s ability to mix data streams with different temporal and spatial resolutions — for example, bringing together single-cell RNA sequencing with time-stamped measurements of serum proteomics, or histopathological imaging. This enables the discovery early predictable markers, and context-dependent therapeutic targets in a mechanistically-informed manner. At its conclusion, this modeling approach will facilitate development of the next generation of precision immunology interventions for real-time monitoring of disease and personalized tide-turning intervention in the treatment of inflammatory diseases. Keywords: Spatiotemporal Modeling; Inflammatory Diseases; Multi-Modal Analytics; Longitudinal Datasets; Immune Dynamics; Precision Immunology 1. Introduction 1.1. Background and Clinical Significance of Inflammatory Diseases Inflammation diseases include a wide-range of diseases with tissue destruction mediated by the immune system, chronic inflammation and deregulated immune responses. These include autoimmune diseases, such as autoimmune diseases as rheumatoid arthritis (RA), IBD, psoriasis, and systemic lupus erythematosus (SLE), which have highly variable clinical courses and organ-specific presentations. On a worldwide scale, hundreds of millions of people suffer from inflammatory conditions that exert enormous social, economic, and health care burdens because of their chronic nature, disability burden, and necessity for long-evoked coping strategies (1). The immune system role in immune-mediated diseases is complex and dynamic. It is a well-organised series of cellular contacts, cytokine signaling and tissue response that depends on temporal parameters and on anatomical locations. In World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 197 normal conditions, inflammation is a defensive response, but in chronic conditions, it becomes pathological leading to tissue damage and systemic manifestations (2). The development of biologics and targeted therapies has opened up new landscapes for treatment, but relies upon timely intervention, accurate patient stratification and immune monitoring at the time of intervention (3). New high throughput techniques (eg, single cell sequencing, spatial transcriptomics and mass cytometry) have broadened our knowledge of the immune landscape. Such powerful instruments discover the remarkable diversity of immune cell, and delineate the spatial and temporal transformations that drive disease pathogenesis (4). However, integration of such knowledge to yield clinical insights has been elusive, in part because data are fragmentary and integrative models that represent immune activity across tissue compartments and over time are lacking. Systemslevel effort is required to connect immune profiling to actionable decision-making in inflammatory disease care (5). 1.2. Challenges in Modeling Immune System Dynamics The modeling of immune-system dynamics is challenging for several reasons, including system complexity, variation, and dimensionality. By contrast to static measures, immune responses are dynamic with rapidly changing cell populations, oscillating levels of cytokines and transient encounters that occur over both temporal and spatial dimensions. Understanding this complex relationship necessitates longitudinal data, spatial resolution, and computational approaches that can detect global signals as well as local disturbances (6). The heterogeneity of immune cell states and the plasticity of their functions represent one of the main barriers as shown in figure 1. A single type of cell can have multiple phenotypes depending on conditions, anatomical location, and exposure to stimuli, and deterministic models are therefore likely to be unsuitable to predict their dynamics (7). Moreover, clinical datasets are usually sparse, asynchronously sampled, and inter-patient heterogeneous, which makes temporal alignment difficult and damages the reliability of models (8). Moreover, multi-omic, imaging, and clinical records integration is technically and analytically challenging. Every data type possess a specific resolution, frequency and bias phenomena that have to be normalized, harmonized and dimensionality reduced while maintaining the biological meaning. Finally, there remains a disconnect between mechanistic models that describe the immune and machine learning models that predict outcomes without interpretability (9). It is important to overcome these challenges to fabricate immunologically relevant, clinically translatable platforms. 1.3. Scope, Objectives, and Article Structure This article provides a general framework for spatiotemporal immune dynamics modeling in inflammatory diseases using integrative multi-modal analytics and high-dimensional time series datasets. It aims to transcend the conceptual and technical divides of existing immune monitoring methods to integrate biological understanding, computational modeling, and clinical observational data within a single analytical framework. Key Challenges I: We strive to show how multiscale modeling approaches may be used to predict, interpret, and modulate immune responses in the context of complicated diseases (10). The paper is organized as follows: Section 2 reviews key elements of the theoretical background on immune dynamics and spatial heterogeneity. Section 3 provides an overview of data acquisition and integration and presents preprocessing methodologies. 4 presents computational modeling methods and verification structures. Applications to real-life under selected inflammatory diseases are described in Section 5. Translational and clinical relevance are discussed in Section 6. Methodological limitations and future research directions are presented in Section 7. The final Section 8 wraps up with a discussion of overall findings and a vision for future application in precision immunology. World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 198 Figure 1 Overview diagram showing the complexity of spatiotemporal immune responses across organs [18] 2. Theoretical framework and biological rationale 2.1. Immune System Architecture in Spatiotemporal Context The immune system is a distributed and multiscale network spaced and meantime that is set up to preserve host defense as well as tissue homeostasis. Its structure encompasses central immune organs (i.e., bone marrow and thymus) and peripheral sites (e.g., lymph nodes, mucosal surfaces, barrier tissues). Immune cells are in constant transit between these different locales, traveling through the blood and lymphatics and reacting to an ever-changing series of internal and external cues, instantly updated (6). This constant motion is the foundation for immune surveillance, tissue healing, and resolution of inflammation. One important feature of immune activity is its context-dependence: the same immune cells can have different behaviour depending on tissue-specific cues, local cytokine concentrations and interactions with resident stromal or epithelial cells. For example, at home alone macrophages have different transcriptional program in lung, liver or skin, which is referred to tissue imprinting (7). Such spatially tagged activity is additionally modulated as a function of time, in emerging and resolving states of immune challenge. Furthermore, lymphoid organs including spleen and lymph nodes provide central hubs for antigen presentation, connecting innate and adaptive immunity that shapes adaptive immune responses. On infection or injury, immune cells are recruited in a temporally coordinated order initial neutrophils, followed by monocytes, dendritic cells and finally adaptive effectors as T and B cells (8). These spatiotemporal cascades are tightly controlled to prevent tissue damage, over inflammation. Under the condition of inflammatory diseases, this spatial cross talk is lost, the immune cells are accumulated at ectopic locations, the barrier function is disrupted and the tissue damage becomes chronic. As such, a detailed portrayal of immune system World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 199 architecture over the space and time is pivotal for the construction of accurate, mechanistic models for disease evolution and therapeutic response (9). 2.2. Temporal Evolution of Immune Cell States The immune response is dynamically evolving in time and time-restrained transitions occur in the state, phenotype and functionality of the cells. The fate of immune cells is dynamically influenced by exposure to antigens, cytokines, costimulatory molecules, and metabolic perturbations. These effects induces transient and persistent changes in gene expression, protein modification, and cell response, which enable immune cells to conform to their environment and the stage of the immune response (10). For instance, naive T cells, when triggered, swiftly proliferate and differentiate into multiple effector subsets, including Th1, Th17, and regulatory T cells, each having unique roles in inflammation and tissue repair. This is partially reversible, as memory cells can re-activate upon re-exposure to antigen. B cells also experience classswitching and affinity maturation in germinal centers before transitioning into long-lived plasma cells or memory B cells (11). The innate immune cells are also plastic, as it adapts over time. Macrophages are able to polarize from a proinflammatory (M1-like) to an anti-inflammatory (M2-like) phenotype in response to stimuli from their microenvironment. This fluidity is not binary but rather a continuum of functional states set by temporal cues (12). To capture these transitions, a longitudinal profiling (preferably at single-cell resolution) is essential to prevent missing the transient states and misclassification of the intermediate phenotypes. This is particularly important in inflammatory diseases in which immune profiles may dramatically differ from a flare to a remission. Ensemble dynamics like these need to be addressed for predictive modeling and intervention timing (13). 2.3. Tissue Microenvironments and Spatial Heterogeneity in Inflammatory Conditions The tissue microenvironment is critical to control immune responses through spatially segregated molecular and cellular milieus. Every tissue contains stromal cells, extracellular matrix elements, and inflammatory architectures that communicate to direct the outcome of an immune response. These factors yield spatial heterogeneity which is a hallmark of both normal and pathologic tissue. This heterogeneity becomes further accentuated in inflammation, wherein immune cell infiltration, vasculature remodeling and fibrosis reprogram the tissue landscape (14). This compartmentalisation of the immune response may occur in diseases such as inflammatory bowel disease (IBD) and rheumatoid arthritis (RA). In IBD, immune cells aggregate within the lamina propria and crypt base to form regional centres of cytokine production, epithelial erosion, and the accumulation of neutrophils (15). In RA, synovial tissue becomes hyperplastic and tertiary lymphoid structures develop in the tissue around blood vessels, which are spatial gradients of B and T cells populations (16). Furthermore, spatial transcriptomics and imaging mass cytometry studies have demonstrated that even in a homogeneous tissue section, expression of immune effector molecules varies greatly. For example, pro-inflammatory cytokines that compromise the local T cells, such as IL-6 and TNF-α, could be confined to isolated niches, and the rest would be relatively immune quiescent. This mosaic signaling does not reflect a homogeneous immune state as perceived in tissue biopsies (17). Drug delivery and therapeutic efficacy are also influenced due to spatial heterogeneity. Fibrosis-dense areas or vascular occlusion could impede drug penetrance, and facilitate resistance to therapy. Consequently, including the spatial context in immune modeling allows better informing disease mechanisms and enhancing the predictive capability of computational simulations in inflammation-related diseases (18). 2.4. Limitations of Conventional Immune Monitoring Techniques Despite the growing accessibility of high-throughput technologies, standard immune monitoring tools are constrained in their capacity to capture the complete complexity of spatiotemporal immune dynamics. Staples of clinical immunology, such as flow cytometry, ELISA and bulk transcriptomics, provide sider to those like snapshots from afar of immune activity, without either spatial resolution or temporal continuity. These techniques are commonly based on sampling of peripheral blood and this does not necessarily reflect immune processes in the target tissue, such as the synovium or gut mucosa (19). In addition, cross-sectional sampling strategies typically employed in clinical trials do not reflect the dynamic changes of immune cell states, and may obscure potential transition or context specific responses. Bulk measurements further mask cell-to-cell heterogeneity that is important for heterogeneity in immune trajectories. Thus, a multitude of cellular World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 200 interactions and microenvironmental stimulations are unobserved, which constrains the granularity in mechanistic models (20). To address these issues, we need new methodologies that combine time-series, spatially resolved data into coherent models that can make sense of complexity, variability and clinical relevance in management of inflammatory disease. This is shown in table 1 below. Table 1 Comparison of Classical vs. High-Dimensional Immune Profiling Technologies Across Spatial and Temporal Scales Dimension Classical Immune Profiling High-Dimensional Immune Profiling Reference Temporal Resolution Low (periodic, point-in-time testing) High (longitudinal, real-time sampling) (85) Spatial Resolution Poor (bulk samples, systemiclevel) High (cellular, tissue-microenvironment scale) (86) Data Dimensionality Limited (few markers, ~10–20) Extensive (100s–1000s of genes, proteins, or spatial pixels) (87) Throughput Medium High (massive parallel profiling of cells or transcripts) (88) Analytical Complexity Low (manual gating, univariate stats) High (machine learning, nonlinear modeling, network inference) (89) Interpretability High (clinician-friendly readouts, standardized assays) Moderate to low (requires computational expertise, visual tools) (90) Clinical Integration Routine (ELISA, flow cytometry) Emerging (scRNA-seq, CyTOF, ST) (81) Cost and Accessibility Low to moderate High (equipment, computational pipelines, licensing) (72) 3. Data sources and multi-modal integration strategy 3.1. Types of Longitudinal and Spatially Resolved Datasets Fine-grained immunological modeling of inflammatory diseases thus requires data sets that are multi-scale, including molecular, cellular, spatial, and temporal information. One of the most revolutionary ones would be single-cell RNA sequencing (scRNA-seq), cytometry by time-of-flight (CyTOF), spatial transcriptomics (ST) and high-throughput imaging. Each of these views has unique strengths in capturing elements of the complexity of immunity, and when combined they provide a strong base from which to model disease progression. 11) The technique of scRNA-seq further makes it possible to perform transcriptomic characterisation of individual immune cells providing information about their expression heterogeneity to assign cell types and state classification. It can be used to identify rare populations, lineage trajectories, or cell fate transitions occurring during inflammation, or in response to therapeutic intervention. Although confined to single cell suspensions, its temporal immune profiling resolution is without equal. CyTOF not only provides single-cell resolution, but also can meaure more than 40 proteins simultaneously with antibodies labeled with heavy metals. It is particularly well suited to measure surface and intracellular markers in a high-throughput multiplexed manner, e.g., for immune phenotyping in longitudinal samples like peripheral blood (12). In the setting of immunotherapy or flare-remission cycles, CyTOF has been crucial in monitoring dynamic immune signatures. Spatial transcriptomics (ST) and Imaging Mass Cytometry (IMC) are also essential for in situ immune responses. ST enables retention of spatial context and transcriptomic data on a genome-wide scale, thus researchers can directly map gene expression onto to tissue histology (13). In contrast, IMC can resolve spatial proteomic distributions at subcellular resolution to analyze cell neighborhoods and niche-specific functions in tissue microenvironments. High-Resolution Multimodal Imaging to Augment Spatial Understanding by Superimposing Architectural and Immunologic Features In addition, high-resolution multimodal imaging, specifically in-room multiplexed immunoflu and confocal microscopy, provides overlay of structural and immunologic features (14). World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 201 However, with their scale and heterogeneity, such data are imperative for linking static, localized pathology and realtime immune kinetics. When integrated across time points, these methods also allow for a mapping of the complete inflammatory immune landscape. 3.2. Temporal Data Streams: Time-Series Proteomics, Cytokine Panels, Clinical Trajectories Time-course data streams play a fundamental role in modeling immune system activity in inflammatory diseases, owing particularly to their nonlinear evolution and episodic dynamics. Time-series proteomics is a class of time-series data source which represents dynamic protein expression measurements at multiple time steps. Techniques such as data-independent acquisition mass spectrometry facilitate the quantification of hundreds of proteins, providing insight into developing immune pathways, metabolic changes, and acute-phase responses (15). Cytokine and chemokine panels, frequently assayed using ELISA, Luminex, or SIMOA, yield high frequency, low volume analyses of intercellular signaling molecules that drive inflammation. These panels are especially useful to investigate paracrine communication and global immune states upon disease exacerbation and resolution (16). Chronologically aligned profiles of cytokines found to occur before clinical flares point to them being "early-warning" signals. From the literature, clinical pathways (e.g., longitudinal patient records, imaging reports, symptom scores and medication histories) providing context for molecular and cell level data have been developed. When appropriately timestamped and synced with medical-record (biological) sampling, such datasets can provide useful information on association between immune activity and clinical subjects such as relapse, remission or adverse events (17). In addition, wearable and digital monitoring devices (e.g., smartwatches and biosensors) are novel tools that produce real-time, continuous physiological measures like heart rate variability (HRV) and parameters of sleep that are frequently associated with systemic inflammation. These signals can be integrated with omic data to inform digital phenotyping approaches and improve temporal resolution in immune models (18). Second, as a whole, the temporal stream of data form the support for studying trajectory inference, dynamic network analysis, and predictive modeling. Their integration with spatial datasets and molecular characteristics contributes to the integral perception of inflammatory immune dynamics in time. 3.3. Spatial Data Streams: Imaging Mass Cytometry, Spatial Transcriptomics Spatial trajectories of immune cells through tissue landscapes are valuable data streams that provide insights into their context-specific behaviors. (11, 12) Together, imaging mass cytometry (IMC) and spatial transcriptomics (ST) are front-line technologies in deciphering the microanatomic landscape of immune responses in inflammatory diseases. IMC combines metal-labeled antibodies and laser ablation with detection, quantification, and identification using TOF mass spectrometry to create high-dimensional proteomic maps of tissue sections at a near-cellular scale. It allows for multiplexed (30–40 proteins simultaneously) quantification preserving single-cell localization. This is especially important in pinpointing cellular neighborhoods, immune niches and structural perturbations in diseased tissues, like synovial pannus or inflamed colonic mucosa (19). By defining immune cell-stromal or immune cell-vascular interactions, IMC provides a functional map of inflammation. Spatial transcriptomics (ST) adds to this by offering whole-transcriptome data localized in space within a section of tissue. Technologies, including 10× Genomics Visium and NanoString GeoMx, allow quantification of mRNA molecules across hundreds to thousands of spatially barcoded spots (20). ST permits mapping gene-expression gradients, discovering spatially-restricted cell states, and reconstructing inflammatory cascades while they occur across anatomical compartments. Both IMC and ST may be added to classical histopathology and immunohistochemistry to form multimodal, interpretable surveys of the tissue immune profile. Additionally, modern registration techniques permit co-mapping of proteomic and transcriptomic layers, thus providing genuine multimodal spatial integration. These techniques will be critical to mapping context-dependent immune phenotypes that are not reflected in the PB or homogenized bulk tissue. Their spatiotemporal resolution is critical for a biologically realistic modeling of immune heterogeneity, cellular communication and therapeutic microenvironments. 3.4. Integration Architecture: Harmonizing Resolution and Timepoints Fusing multi-modal data streams at various spatial and temporal resolutions is a challenging yet indispensable aspect of spatiotemporal immune modeling. It is in this context that a flexible but structured integration architecture is World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 202 needed to map and homogenize heterogeneous data sources into a unified analytical universe. Its architecture is usually composed of three layers: data ingestion, intermediate alignment, and joint embedding. Data from datasets like scRNA-seq, CyTOF, ST, and clinical records are extracted and linked using shared concepts, such as patient ID, tissue type, and time point, during ingestion. But source data come in different depths, frequencies, and resolutions, requiring preprocessing and metadata enhancement. Temporal snapshots are aligned to spatial coordinates by anchor points, such as patterns of biomarker expression, imaging landmarks, or clinical events. Techniques such as canonical correlation analysis, manifold alignment, variational autoencoders are employed to coregister data from different modalities (21). Finally, the united representation layer projects high-dimensional data into coherent embeddings that captures crucial biological variability without sacrificing the capability for cross-modality comparisons. This step is crucial to further modeling tasks – such as clustering, classification, or trajectory inference Downstream analysis and extension After the above two steps, the filtering is completed. Interoperability and Reproducibility in Integrative Studies Our modular design allows scalability, reproducibility, and interoperability, all required for deploying integrative pipelines in the research and clinical arenas. It provides a platform for comprehensive modeling of the immune system. 3.5. Preprocessing and Data Harmonization Strategies Prior to integration and modelling, multi-modal datasets require extensive pre-processing and harmonisation to remove noise and inconsistencies, and to homogenise the structure. These procedures are essential to maintain biological relevancy and control for bias from experimental variance or technical noise. Preprocessing of scRNA-seq and ST data involves quality control (e.g., mitochondrial genes), normalization (e.g., by SCTransform or by log normalization), and feature selection (e.g., high variable genes). Batch-effects are corrected for using batch-correction tools like Harmony, MNN or Seurat's integration (22). For CyTOF and IMC data-sets, preprocessing includes compensation, signal normalization, and bead-based calibration. Data distribution transformation using arcsinh or logicle is frequently employed to compensate for intensity distribution heavy-tailedness. To accurately demarcate cell boundaries, trained tools or deep learning algorithms (e.g., Ilastik or Cellpose) are employed to segment cells from imaging data. The clinical records and the proteomic datasets must be synchronized in time, the missing values should be estimated, and the units should be standardized. Methods for temporal smoothing, such as Kalman filtering, can be used to compensate for missing data or variability in cytokine panels or wearable sensor streams. Finally, feature scaling and dimensionality reduction like PCA/UMAP are also performed across modalities to ensure the features are compatible for downstream analysis. Harmonization controls the next models to learn biologically meaningful patterns, as opposed to technical variation as show in figure 2. Figure 2 Multi-modal data integration pipeline with temporal and spatial harmonization layers World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 203 Table 2 Summary of Public and Proprietary Multi-Modal Datasets with Immune Relevance Dataset Name Modality Types Disease Focus Resolution Access Type Key Features Reference (Vancouver ) HCA Immune Census scRNA-seq, CyTOF, CITEseq Healthy & Inflammatory tissues Single-cell, multitissue Public Standardized immune profiling across tissues (87) ImmPort Flow cytometry, microarray, cytokine panels Autoimmune diseases, vaccine response Cohortlevel, longitudina l Public Curated immunologica l clinical trials (88) Lupus Nephritis Atlas scRNA-seq, spatial transcriptomic s Systemic Lupus Erythematosu s Cell-type & spatial Public Multi-organ SLE atlas from AMP initiative (72) RA Synovium Atlas scRNA-seq, ATAC-seq, Imaging Mass Cytometry Rheumatoid Arthritis Tissuespecific Public Integrated RA joint and blood immune datasets (73) 10x Genomics Datasets Spatial transcriptomic s, scRNA-seq RA, IBD, SLE, oncology Subcellular to tissuelevel Public/Privat e Visium samples, tissueresolved immune profiling (89) Singlera ImmunoBank Single-cell DNA methylation + transcriptomic s Lupus, Crohn’s Disease Single-cell, longitudina l Proprietary Epigenetic lineage tracing with immuno-state mapping (90) FoundationOne Heme Targeted NGS, immune profiling Hematologic malignancies Bulk, targeted panels Proprietary Mutational + immune biomarker landscape for blood cancers (87) Tempus xT + ImmunoINSIGH T RNA-seq, whole-exome sequencing, proteomics Oncology, Inflammatory Disease Mixed (bulk and single-cell) Proprietary AI-augmented clinical decision datasets (77) 4. Computational modeling of spatiotemporal immune dynamics 4.1. Tensor-Based Decomposition of High-Dimensional Longitudinal Data High-dimensional longitudinal data generated by immunological studies (for example, time-resolved single-cell RNA sequencing (scRNA-seq), multiplex cytokine panels, and imaging mass cytometry) are challenging to analyze because they have complex data structures due to innate temporal, spatial, and cellular heterogeneity. Decomposition methods for such multi-dimensional arrays are known as a tensor-based decomposition which gives a natural and mathematically rigorous way to deal with them. World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 204 In contrast to classical matrix approaches, tensors enable the joint modelling of time, modality and, sample or cell identity (15). The most popular tensor analysis approach in this context is Canonical Polyadic (CP) decomposition, which breaks the tensor down to a sum of rank-one components and uncovers patterns sharing by the dimensions (16). For example, a three-mode tensor representing “genes × timepoints × conditions” is decomposed into interpretable vectors that capture temporal gene expression modules associated to disease progression or treatment response. The CP decomposition works well to identify patterns in immune activation and relates them to bouts of disease such as lupus or Crohn’s disease. Besides CP, Tucker decomposition is frequently used for denoising, core feature extraction, and retaining the hierarchies. This is very helpful in multi-batch clinical valuable datasets where biological signal should along patients be preserved (17). More recent developments also include sparsity penalties and non-negative constraints, which, by enforcing interpretation-friendly factor loading matrices, add interpretability to solutions. Tensor-based methods have been integrated with neural networks in hybrid models that have increased the prediction accuracy of inflammatory episodes using features of previous structure in the immune response. In addition, their versatility in diminishing the effect of missing data is preferable for clinical health and NGS data, which have partial observations (18). These may be valuable tools for dimensionality reduction, biomarker discovery, and for revealing temporal signatures in high-throughput longitudinal studies. 4.2. Dynamic Bayesian Networks for Cell-State Transitions Dynamic Bayesian Networks (DBNs) offer a probabilistic graphical approach to describing and learning the temporal relationships and progression of the states of the immune cells across time. Well, in contrast to static Bayesian networks, which codify the conditional dependency of events at one time, DBNs take time series measurements and infer the casual structures over multiple steps in cellular differentiation or activation (19). Immune cell states in inflammation achieve nonlinear dynamics driven by cytokines, antigen exposure, or drugs. A DBN represents these transitions as a sequence of connected states with each node representing a molecular marker or a phenotypic trait and edges showing the probabilistic influence between time slices. This facilitates discovery at the level of causal regulatory circuits – for instance, understanding on how TNF-α stimulation activates NF-κB that in turn leads to monocyte polarization (20). One advantage of DBNs is that they can handle both continuous (e.g., gene expression) and discrete (e.g., cell type) data. Their ability to model latent prior variables would be particularly important (since something like CD4+ T-cell exhaustion (a simulated outcome variable) may not be directly measurable, but rather extrapolated from surrogate data (in our case, CD4+ T-cell exhaustion as a function of checkpoint molecule expression trajectories) in systems immunology) (21). There are a number of learning algorithms for DBN structure learning, such as Expectation-Maximisation and scorebased algorithms. These approaches are frequently combined with bootstrapping to improve robustness and to combat overfitting, particularly when dealing with limited timepoints or the sparse coverage of the cell population. Recent deep-learning-assisted DBNs use RNNs as a prior to speed up convergence for noisy data, a type of data from experimental platforms such as CyTOF or ST (22). DBNs have been successfully used to map immune reconstitution following transplantation, to monitor vaccine-induced immunity, and to discover relapse predictors in autoimmunity. Their interpretability and capacity to represent uncertainty make them appealing for clinical translation, for example to simulate the manner in which therapeutic blockade of IL-6 may change the preferredness of dendritic cell trajectories to go in the direction of tolerogenic phenotypes (23). Together, our DBNs represent a crucial link between empirical data and mechanistic understanding, opening opportunities for hypothesis testing around dynamic immune system changes over time both in experimental and clinical (low-resource emerging) settings. 4.3. Manifold Learning for Temporal Embedding Manifold learning methods are a class of non-linear dimensionality reduction methods which aim to shift highdimensional immune data to smaller low-dimensional manifolds, maintaining the structure of the original data in the process. This is particularly useful in modeling time series data such as when the immune system is transitioning slowly and possibly non-linearly through various states, as in early inflammation to chronic immune dysregulation (24). World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 211 context of spatial transcriptomics and high resolution pathology, will necessitate new governance models for consent, de-identification and secure data transfer (69). At an operational level, to implement spatiotemporal modeling in the context of existing clinical protocols requires coordination across multiple stakeholders—from pathologists and immunologists to IT personnel and reimbursement agencies. Pilot work at tertiary rheumatology and gastroenterology practices is indicating feasibility and value in minimizing diagnostic delays and enhancing patient outcomes. Despite the challenges, continued improvements in modeling, miniaturization of hardware, cloud-based analytics, and policy change are rapidly closing the translational gap. Spatiotemporal immune analytics in personalized medicine is no longer hypothetical—it is becoming a clinical necessity. Figure 5 Conceptual flow from data acquisition to clinical decision support through spatiotemporal modeling 7. Limitations and future research directions 7.1. Data Sparsity, Heterogeneity, and Missingness in Longitudinal Data Multi-omics and clinical longitudinal immune profiles are frequently incomplete, non-uniform, and sparse, thus posing crucial challenges for downstream modeling. Losses of data can be induced by incomparable sampling times, technical dropouts in single-cell assays and patients’ dropout in longitudinal cohorts, resulting in biased estimates and reduced generalizability of prediction models (70). Additionally, different platforms, e.g., scRNA-seq, CyTOF, and proteomics, have diverse noise profiles and dynamic ranges, further complicating the heterogeneity. These problems have been approached through techniques like matrix factorization, autoencoders and variational inference models. However, the latter will tend to introduce artificial signals or strengthen batch effects when extended to sparse matrices (71). Integration harmonization methods such as MOFA+ and Scanorama address this to some extent by learning shared Latent spaces, but are inherently constrained when dealing with nonaligned time series or discontinuous spatial coverage (72). There is also a weakness in inter-rater variability. The immune cell state space is World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 212 dynamic over time, influenced by environmental, genetic, and therapeutic factors, which necessitates hierarchical model-based strategies to separate the within-patient trajectory from population-level trends (73). Without proper normalization, inferred transitions or clusters may be, to a large extent, sampling noise or technical artifacts as opposed to true biological processes. Accordingly, further development of longitudinal immune modeling will require statistical methodologies able to handle temporal discontinuities, reconcile cross-platform heterogeneity, and account for missing data without risk of overfitting or propagation of bias (81). 7.2. Technical Limitations in Spatial Resolution and Cell Identification Even with substantial progress made in spatial omics, resolution and cell identity are still major obstacles. Tools such as imaging mass cytometry (IMC) and spatial transcriptomics (ST) can deliver high-dimensional information on cell localization, but may not attain single-cell resolution on large tissue regions (74). This limits the ability to trace the rare immune cell populations or separate closely intermingled cellular niches. Additionally, existing segmentation algorithms that infer cell boundaries from multiplexed images fail in inflamed or fibrotic tissue where tissue structure is distorted. Segmentation mistakes affect downstream processing, biasing neighborhood analysis and niche-level interaction modeling (75). One can also hypothesize that marker panels used in spatial profiling can be insufficient for the discrimination of overlapping immune phenotypes such as activated DCs vs Tissue-resident macrophages (82). Incorrect labeling has implications on trajectory construction and niche mapping, especially in diseases with a dynamical disease state. To conquer these bottlenecks, better multiplexing, hybrid imaging genomics, and AI powered image analysis are required (76). 7.3. Ethical and Governance Challenges in Integrating Multi-Source Patient Data Multi-omic, imaging, clinical records, and wearables integration Multi-modal data integration - integrating genomics, imaging, clinical records and wearables - poses significant ethical and governance challenges. With spatial and temporal elements, de-identification becomes progressively harder, particularly if connected to specific tissue coordinates or digital health logs (77). This increases the likelihood of re-identification and accidental data disclosure. Furthermore, existing consent models seldom consider that multi-modal datasets may be re-used in multiple studies or across institutions. Patients may agree to a clinical biopsy, but not to downstream use for spatial transcriptomic profiling or to machine learning training (78). To the best of our knowledge, regulatory body has yet to define a solid data governance standard balancing utility and privacy for such high-resolution biomedical data (83). Biases inherent in training data whether they are encoded by underrepresentation of minority groups or geographic subsets can also contribute to susceptibility to algorithmic discrimination in precision medicine use cases. The importance of establishing federated learning, differential privacy, and rigorous ethical review mechanisms should not be underestimated in order to protect patient autonomy while promoting scientific advancement (79). 7.4. Prospects for Real-Time, In Situ Immune Monitoring Systems New biosensing systems coming online offer the potential for in situ real-time immune monitoring at the tissue interface. 80) Wearable or implantable devices incorporating microfluidic chips and electrochemical immunosensors can take real-time measurements of cytokines dynamics, immune cell activation, and metabolic signals (80). When combined with AI-driven interpretation engines and digital dashboards, these tools facilitate “closed-loop” decision support, wherein therapy is modulated in real time to accommodate shifts in immunity (84). Although in the early stages of development, these systems are anticipated for early flare detection, remote patient monitoring, and adaptive immunotherapy as much as they are revolutionizing the management of chronic inflammation by being precise, autonomous with minimal invasiveness. World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 213 8. Conclusion 8.1. Summary of Key Findings and Framework Contributions This paper has offered an in-depth study about the representation of spatiotemporal immune dynamics in inflammatory diseases using integrative multi-modal analytics and high-dimensional longitudinal data. Our ‘fused phenome-immunome’ framework integrating multimodal immune data ranging from single-cell transcriptomics and spatial proteomics to clinical trajectories and digital health inputs allows a more comprehensive, temporally-informed, and spatially-resolved understanding of immune dysregulation across diseases including rheumatoid arthritis, inflammatory bowel disease, and systemic lupus erythematosus. A key aspect to our approach is the framing of a mathematical framework which unifies data across resolution and time, addressing the challenge of fragmented immune snapshots, which has been uncoupled for decades. We demonstrated the use of tensor decomposition, dynamic Bayesian networks, and manifold learning to reconstruct immune cell trajectories and infer future states of inflammation. This provides a descriptive language for pathophysiology and a predictive machine for clinician decision making at the point of care. Our results also highlight the importance of spatial modeling-ensemble using graph-based statistics and niche interaction networks-to understand immune responses in tissue microenvironments. Combining these spatial maps with temporal pathways extends interpretability and permits inferences that are informed by both when and where immune changes take place. Importantly, we demonstrated several high value use cases, such as early biomarker discovery, precise therapy stratification, and real-time flare prediction. The ability to translate immune complexity into clinically useful outputs – in the form of dashboards, predictive alerts, and individualized risk stratification – signifies a revolution in the future monitoring and management of inflammatory disease. In our disease-focused case studies, we demonstrated that immune models may overcome the silos imposed by disease, and give rise to shared pathways and mechanisms, thus supporting cross-disease therapeutic intervention. This integrative modeling will change the way in which immune-mediated diseases are understood - from static classification to dynamic, multi-dimensional representation. 8.2. Strategic Vision for Future Integrative Immune Modeling in Clinical Settings On the horizon, the strategic use of spatiotemporal immune modeling within clinical practice will require more than the adoption of new technologies; it will necessitate system-level change embracing infrastructure, policy, and education. To achieve this vision, we need to rethink hospitals and research centers as “living immune observatories,” enabling data to be integrated rapidly and fluidly across time points, modalities, and departments. One such approach would be to incorporate AI-empowered modeling tools within electronic health records and diagnostic platforms to enable real-time immune status dashboards. Such systems will raise alarms for subclinical changes, categorize the patient into archetype of clinical trajectory, and suggest the best window for interventions. For that to happen, entities will need to invest in interoperable digital health ecosystems and cross-platform standardization. Under policy, regulatory frameworks will have to learn to cope with dynamic biomarker validation, longitudinal data governance, and model update cycles. Institutional review boards/regulatory agencies will need to design clear guidelines for the employment of adaptive algorithms for immune monitoring to maintain ethical integrity and patient confid1ence. It is also important to have clinician education. Making sense of high-dimensional immune outputs will demand new fluencies in data literacy, systems immunology, and computational diagnostics. Embedment of these skills in medical curriculum and/or CME will be also necessary to drive acceptance. And finally… equity must continue to be centered in implementation. Models need to be stress-tested in different patient populations, settings of care, socioeconomic strata, and global geographic settings in order to avoid algorithmic bias and to guarantee that all segments of the population are served well. The path forward is arduous but hopeful. Through joint efforts involving science, medicine, technology and policy, spatiotemporal immune modeling can transition from a research frontier to a foundation of precision inflammatory disease control. It has the power to turn immunologic cacophony into clarity—illuminating therapies, predicting disease, and tailoring care with a level of granularity not previously possible. World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 214 Compliance with ethical standards Disclosure of conflict of interest No conflict of interest to be disclosed. References [1] Firestein Gary S, Budd Ralph C, Gabriel Sherine E, McInnes Iain B, O'Dell James R. Kelley's Textbook of Rheumatology. 10th ed. Philadelphia: Elsevier; 2017. [2] Chen Lanfang, Deng Hua, Cui Hong. The inflammatory response and its regulation in systemic lupus erythematosus. Journal of Autoimmunity. 2020;113:102589. doi: 10.1016/j.jaut.2020.102589 [3] Smolen Josef S, Aletaha Daniel, McInnes Iain B. Rheumatoid arthritis. The Lancet. 2016;388(10055):2023–38. doi: 10.1016/S0140-6736(16)30173-8 [4] Stubbington Matthew J T, Rozenblatt-Rosen Orit, Regev Aviv, Teichmann Sarah A. Single-cell transcriptomics to explore the immune system in health and disease. Science. 2017;358(6359):58–63. doi: 10.1126/science.aan6828 [5] Farh Keith K-H, Marson Alexander, Zhu Jun, Kleinewietfeld Markus, Housley William J, Beik Shermin, Shoresh Noam, Whitton Howard, Ryan Robert JH, Shishkin Alexander A, Hatan M. Genetic and epigenetic fine mapping of causal autoimmune disease variants. Nature. 2015;518(7539):337–43. doi: 10.1038/nature13835 [6] Satija Rahul, Farrell Jeffrey A, Gennert David, Schier Alexander F, Regev Aviv. Spatial reconstruction of single-cell gene expression data. Nature Biotechnology. 2015;33(5):495–502. doi: 10.1038/nbt.3192 [7] Papalexi Efthymia, Satija Rahul. Single-cell RNA sequencing to explore immune cell heterogeneity. Nature Reviews Immunology. 2018;18(1):35–45. doi: 10.1038/nri.2017.76 [8] Hochreiter Sepp, Clevert Djork-Arné, Obermayer Klaus. A new summarization method for single-cell RNA-seq data. Bioinformatics. 2018;34(18):3115–20. doi: 10.1093/bioinformatics/bty175 [9] Yu Kevin H, Beam Andrew L, Kohane Isaac S. Artificial intelligence in healthcare. Nature Biomedical Engineering. 2018;2(10):719–31. doi: 10.1038/s41551-018-0305-z [10] Regev Aviv, Teichmann Sarah A, Lander Eric S, Amit Ido, Benoist Christophe, Birney Ewan, Bodenmiller Bernd, Campbell Peter, Carninci Piero, Clatworthy Menna, Clevers Hans. The Human Cell Atlas. eLife. 2017;6:e27041. doi: 10.7554/eLife.27041 [11] Mowat Allan M, Agace William W. Regional specialization within the intestinal immune system. Nature Reviews Immunology. 2014;14(10):667–85. doi: 10.1038/nri3738 [12] Ginhoux Florent, Guilliams Martin. Tissue-resident macrophage ontogeny and homeostasis. Immunity. 2016;44(3):439–49. doi: 10.1016/j.immuni.2016.02.024 [13] Iwasaki Akiko, Medzhitov Ruslan. Control of adaptive immunity by the innate immune system. Nature Immunology. 2015;16(4):343–53. doi: 10.1038/ni.3123 [14] Krummel Matthew F, Bartumeus Ferran, Gérard Alain. T cell migration, search strategies and mechanisms. Nature Reviews Immunology. 2016;16(3):193–201. doi: 10.1038/nri.2015.16 [15] Kaech Susan M, Cui Weiguo. Transcriptional control of effector and memory CD8+ T cell differentiation. Nature Reviews Immunology. 2012;12(11):749–61. doi: 10.1038/nri3307 [16] Victora Gabriel D, Nussenzweig Michel C. Germinal centers. Annual Review of Immunology. 2012;30:429–57. doi: 10.1146/annurev-immunol-020711-075032 [17] Murray Peter J. Macrophage polarization. Annual Review of Physiology. 2017;79:541–66. doi: 10.1146/annurevphysiol-022516-034339 [18] Kazer Samuel W, Aicher Thomas P, Muema Daniel M, Carroll Sean L, Ordovas-Montanes Jose, Miao Victoria N, Tu Anjie A, Ziegler Carly G K, Nyquist Sarah K, Wong Ernest B, Ndung’u Thumbi. Integrated single-cell analysis of multicellular immune dynamics during hyperacute HIV-1 infection. Nature Medicine. 2020;26(4):511–8. doi: 10.1038/s41591-020-0782-5 World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 215 [19] Aran Dvir, Looney Mark R, Liu Zicheng. Analysis of tissue-specific gene expression and splicing from single-cell RNA-seq data. Genome Biology. 2019;20:180. doi: 10.1186/s13059-019-1764-4 [20] Friedrich Markus, Pohin Mathieu, Powrie Fiona. Cytokine networks in the pathophysiology of inflammatory bowel disease. Immunity. 2019;50(4):992–1006. doi: 10.1016/j.immuni.2019.03.017 [21] Weyand Cornelia M, Goronzy Jörg J. Immunometabolism in early and late stages of rheumatoid arthritis. Nature Reviews Rheumatology. 2017;13(5):291–301. doi: 10.1038/nrrheum.2017.37 [22] Schürch Christian M, Bhate Shakti S, Barlow Glenn L, Phillips Daniel J, Noti Luca, Zlobec Inti, Nolan Garry P. Coordinated cellular neighborhoods orchestrate antitumoral immunity at the colorectal cancer invasive front. Cell. 2020;182(5):1341–1359.e19. doi: 10.1016/j.cell.2020.07.005 [23] Wu Tiffany D, Madireddi Srilatha, de Almeida Patrícia E, Banchereau Rémi, Chen Yajing J, Chitre Amit S, Chiang Evelyn Y, Iftikhar Hina, O'Gorman Walter E, Au-Yeung Annie, Takahashi Claire. Peripheral T cell expansion predicts tumour infiltration and clinical response. Nature. 2020;579(7798):274–8. doi: 10.1038/s41586-0202056-8 [24] Mahnke Yvonne D, Brodie Thomas M, Sallusto Federica, Roederer Mario, Lugli Enrico. The who's who of T-cell differentiation: Human memory T-cell subsets. European Journal of Immunology. 2013;43(11):2797–809. doi: 10.1002/eji.201343751 [25] Bjornson Zachary B, Nolan Garry P, Fantl Wendy J. Single-cell mass cytometry for analysis of immune system functional states. Current Opinion in Immunology. 2013;25(4):484–94. doi: 10.1016/j.coi.2013.04.005 [26] Spitzer Matthew H, Nolan Garry P. Mass cytometry: single cells, many features. Cell. 2016;165(4):780–91. doi: 10.1016/j.cell.2016.04.019 [27] Ståhl Patrik L, Salmén Fredrik, Vickovic Sanja, Lundmark Anders, Navarro José F, Magnusson Johan, Giacomello Stefania, Asp Michaela, Westholm Jakub O, Huss Mikael, Mollbrink Annelie. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78–82. doi: 10.1126/science.aaf2403 [28] Lin Jiansheng R, Izar Benjamin, Wang Shanrong, Yapp Chia-Ling, Mei Sheng, Shah Preeti M, Santagata Sandro, Sorger Peter K. Highly multiplexed immunofluorescence imaging of human tissues and tumors using t-CyCIF and conventional optical microscopes. eLife. 2018;7:e31657. doi: 10.7554/eLife.31657 [29] Gillet Ludovic C, Leitner Alexander, Aebersold Ruedi. Mass spectrometry applied to bottom-up proteomics: entering the high-throughput era for hypothesis testing. Annual Review of Analytical Chemistry. 2016;9:449–72. doi: 10.1146/annurev-anchem-071015-041550 [30] Leng Sophie X, McElhaney Janet E, Walston Jeremy D, Xie Dan, Fedarko Neal S, Kuchel George A. ELISA and multiplex technologies for cytokine measurement in inflammation and aging research. The Journals of Gerontology: Series A. 2008;63(8):879–84. doi: 10.1093/gerona/63.8.879 [31] Aletaha Daniel, Smolen Josef S. Diagnosis and management of rheumatoid arthritis: a review. JAMA. 2018;320(13):1360–72. doi: 10.1001/jama.2018.13103 [32] Quer Guillermo, Gouda Payal, Galarnyk Max, Topol Eric J, Steinhubl Steven R. Interand intraindividual variability in daily resting heart rate and its associations with age, sex, sleep, BMI, and time of year: retrospective, longitudinal cohort study of 92,457 adults. PLoS One. 2020;15(2):e0227709. doi: 10.1371/journal.pone.0227709 [33] Giesen Charlotte, Wang Haifang A O, Schapiro Denis, Zivanovic Natasa, Jacobs Andre, Hattendorf Bodo, Schüffler Patrick J, Grolimund Daniel, Buhmann Joachim M, Brandt Sibylle, Ziegler Andrea. Highly multiplexed imaging of tumor tissues with subcellular resolution by mass cytometry. Nature Methods. 2014;11(4):417–22. doi: 10.1038/nmeth.2869 [34] Stickels Robert R, Murray Emily, Kumar Pratik, Li Jie, Marshall John L, Di Bella Dominic J, Arlotta Paola, Macosko Evan Z. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2. Nature Biotechnology. 2021;39(3):313–9. doi: 10.1038/s41587-020-0739-1 [35] Welch Joshua D, Kozareva Veneta, Ferreira Alessandra, Vanderburg Caroline, Martin Casey, Macosko Evan Z. Single-cell multi-omic integration compares and contrasts features of brain cell identity. Cell. 2019;177(7):1873– 87.e17. doi: 10.1016/j.cell.2019.05.006 World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 216 [36] Korsunsky Ilya, Millard Nathaniel, Fan Jean, Slowikowski Kamil, Zhang Fan, Wei Kejun, Baglaenko Yury, Brenner Michael, Loh Po-Ru, Raychaudhuri Soumya. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature Methods. 2019;16(12):1289–96. doi: 10.1038/s41592-019-0619-0 [37] Kolda Tamara G, Bader Brett W. Tensor decompositions and applications. SIAM Review. 2009;51(3):455–500. doi: 10.1137/07070111X [38] Acar Evrim, Dunlavy Daniel M, Kolda Tamara G. A scalable optimization approach for fitting canonical tensor decompositions. Journal of Chemometrics. 2011;25(2):67–86. doi: 10.1002/cem.1335 [39] Hore Valerie, Vieth Beate, Bacher Rhonda, Soneson Charlotte, Robinson Mark D. Tensor decomposition for multiple-tissue gene expression experiments. Nature Methods. 2016;13(3):241–4. doi: 10.1038/nmeth.3768 [40] Sidiropoulos Nicholas D, De Lathauwer Lieven, Fu Xiaodong, Huang Yimin, Papalexakis Evangelos E, Faloutsos Christos. Tensor decomposition for signal processing and machine learning. IEEE Transactions on Signal Processing. 2017;65(13):3551–82. doi: 10.1109/TSP.2017.2690524 [41] Murphy Kevin P. Dynamic Bayesian Networks: Representation, Inference and Learning. UC Berkeley: Department of Computer Science; 2002. [42] Sachs Karen, Perez Oded, Pe’er Dana, Lauffenburger Douglas A, Nolan Garry P. Causal protein-signaling networks derived from multiparameter single-cell data. Science. 2005;308(5721):523–9. doi: 10.1126/science.1105809 [43] Ahmed Sazzad, Lin Fangzhou, Hasegawa Koji, Ghosh Soma, Kitano Hiroaki. Learning cell signaling networks from high throughput proteomic data using dynamic Bayesian network. BMC Bioinformatics. 2019;20(Suppl 6):233. doi: 10.1186/s12859-019-2855-4 [44] Li Bin, Huang Wenqi, Wang Dapeng, Wang Zhiwei. A deep dynamic Bayesian network model for real-time prediction in health care. IEEE Access. 2020;8:152547–57. doi: 10.1109/ACCESS.2020.3017662 [45] Zhang Wei, Yu Rongrong, Zhang Yuanyuan, Li Zhaoyang, Wang Jiajia, Li Xiaoqian, Chen Jing, Zhang Shizhong. Immune reconstitution and dynamic immune network analysis using Bayesian modeling. Frontiers in Immunology. 2021;12:631012. doi: 10.3389/fimmu.2021.631012 [46] Moon Kevin R, van Dijk David, Wang Zhong, Burkhardt Daniel B, Chen Weng, van den Elzen Antonia, Hirn Matthew J, Coifman Ronald R, Ivanova Natalia B, Wolf Guy, Krishnaswamy Smita. Visualizing structure and transitions in high-dimensional biological data. Nature Biotechnology. 2019;37(12):1482–92. doi: 10.1038/s41587-019-0336-3 [47] Haghverdi Laleh, Buettner Florian, Theis Fabian J. Diffusion maps for high-dimensional single-cell analysis of differentiation data. Bioinformatics. 2015;31(18):2989–98. doi: 10.1093/bioinformatics/btv325 [48] Trapnell Cole, Cacchiarelli Davide, Grimsby John, Pokharel Pallavi, Li Shaojie, Morse Michael, Lennon Nicholas J, Livak Kenneth J, Mikkelsen Tarjei S, Rinn John L. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nature Biotechnology. 2014;32(4):381–6. doi: 10.1038/nbt.2859 [49] Wolf Fabian A, Hamey Fiona K, Plass Marvin, Solana Juan, Dahlin Joakim S, Göttgens Berthold, Rajewsky Nikolaus, Simon Linnarsson, Theis Fabian J. PAGA: graph abstraction reconciles clustering with trajectory inference through a topology preserving map of single cells. Genome Biology. 2019;20(1):59. doi: 10.1186/s13059-0191663-x [50] Schürch Christian M, Bhate Shalin S, Barlow Glenn L, Phillips Daniel J, Noti Lukas, Zlobec Inti, Nolan Garry P, Cheng Liang, Bodenmiller Bernd. Coordinated cellular neighborhoods orchestrate antitumoral immunity at the colorectal cancer invasive front. Cell. 2020;182(5):1341–59.e19. doi: 10.1016/j.cell.2020.07.005 [51] Cang Zixuan, Nie Qing. Inferring spatial and signaling relationships between cells from single cell transcriptomic data. Nature Communications. 2020;11:2084. doi: 10.1038/s41467-020-15946-9 [52] Zhao Enwei, Stone Matthew R, Ren Xiyu, Pulliam Tasha, Nghiem Paul, Bielas Jason H, Gottardo Raphael, Huang Chenling Y. Spatial transcriptomics at subspot resolution with BayesSpace. Nature Biotechnology. 2021;39(11):1375–84. doi: 10.1038/s41587-021-00935-2 [53] Zhang Lili, Yu Xin, Zheng Ling, Zhang Yajing, Wang Jingjing, Zhang Zehang, Fang Qin, Gao Ruoyi, Kang Bo, Zhang Nianping, Xiao Feng. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature. 2018;564(7735):268–72. doi: 10.1038/s41586-018-0694-x World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 217 [54] Dries Ruben, Zhu Qingyang, Dong Rui, Eng Cathy-Hua Lin, Li Haoxiang, Liu Kai, Fu Yu-Cheng, Zhao Tsz-Ying, Sarkar Abhik, Bao Fa, George Roy E. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biology. 2021;22(1):78. doi: 10.1186/s13059-021-02286-2 [55] Wang Jiayu, Agarwal Deepak, Huang Jing. Semi-supervised learning for cross-domain recommendation in online social rating networks. IEEE Transactions on Knowledge and Data Engineering. 2019;31(7):1283–96. doi: 10.1109/TKDE.2018.2833447 [56] Liu Qiaolin, Ma Chang, Shen Suwen, Cao Rui, Wang Jiayi, Wang Zhen, Lin Kai. Deep learning-based multimodal fusion for single-cell data. Nature Methods. 2023;20(1):89–98. doi: 10.1038/s41592-022-01647-5 [57] Lotfollahi Mohammad, Wolf Fabian A, Theis Fabian J. scGen predicts single-cell perturbation responses. Nature Methods. 2019;16(8):715–21. doi: 10.1038/s41592-019-0494-8 [58] Ji Zhe, Zhou Weiqiang, Hou Ye, Yuan Qiang, Wang Yuan, Yu Le, Zhang Qian, Jin Jian, Li Weixuan, Liu Ji. Spatial transcriptomic analysis of glioblastoma identifies a vascular niche with novel tumor-associated macrophages. Nature Communications. 2020;11(1):1–14. doi: 10.1038/s41467-020-17661-z [59] Andersson Anna, Bergenstråhle Johan, Asp Michaela, Bergenstråhle Ludvig, Jurek Anna, Fernández Navarro José, Lundeberg Joakim. Single-cell and spatial transcriptomics enables probabilistic inference of cell type topography. Communications Biology. 2020;3(1):1–8. doi: 10.1038/s42003-020-1127-2 [60] Sundararajan Mukund, Taly Ankur, Yan Qiqi. Axiomatic attribution for deep networks. In: International Conference on Machine Learning. PMLR; 2017:3319–28. doi: 10.48550/arXiv.1703.01365 [61] Stephenson W, Donlin LT, Butler A, Rozo C, Bracken B, Rashidfarrokhi A, Goodman SM, Ivashkiv LB, Bykerk VP, Orange DE, Satija R. Single-cell RNA-seq of rheumatoid arthritis synovial tissue using low-cost microfluidic instrumentation. Nature Communications. 2018;9(1):791. doi: 10.1038/s41467-017-02659-x [62] Zhang F, Wei K, Slowikowski K, Fonseka CY, Rao DA, Kelly S, Goodman SM, Tabechian D, Hughes LB, SalomonEscoto K, Watts GFM. Defining inflammatory cell states in rheumatoid arthritis joint synovial tissues by integrating single-cell transcriptomics and mass cytometry. Nature Immunology. 2019;20(7):928–42. doi: 10.1038/s41590-019-0378-1 [63] Rao DA, Gurish MF, Marshall JL, Slowikowski K, Fonseka CY, Liu Y, Donlin LT, Henderson LA, Wei K, Mizoguchi F, Teslovich NC. Pathologically expanded peripheral T helper cell subset drives B cells in rheumatoid arthritis. Nature. 2017;542(7639):110–4. doi: 10.1038/nature20810 [64] Smillie CS, Biton M, Ordovas-Montanes J, Sullivan KM, Burgin G, Graham DB, Herbst RH, Rogel N, Slyper M, Waldman J, Sud M. Intraand inter-cellular rewiring of the human colon during ulcerative colitis. Cell. 2019;178(3):714–30.e22. doi: 10.1016/j.cell.2019.06.029 [65] Martin JC, Chang C, Boschetti G, Ungaro R, Giri M, Grout JA, Gettler K, Chuang LS, Nayar S, Greenstein AJ, Dubinsky M. Single-cell analysis of Crohn’s disease lesions identifies a pathogenic cellular module associated with resistance to anti-TNF therapy. Cell. 2019;178(6):1493–508.e20. doi: 10.1016/j.cell.2019.08.008 [66] Haberman Y, Tickle TL, Dexheimer PJ, Kim MO, Tang D, Karns R, Baldassano RN, Noe JD, Rosh J, Markowitz J, Baker SS. Pediatric Crohn disease patients exhibit specific ileal transcriptome and microbiome signature. The Journal of Clinical Investigation. 2014;124(8):3617–33. doi: 10.1172/JCI75436 [67] Danese S, Vermeire S, Hellstern P, Panaccione R, Rogler G, Tomasello G, D’Haens G, Peyrin-Biroulet L. Randomised trial and open-label extension study of an anti-MAdCAM-1 antibody in ulcerative colitis. Gut. 2018;67(6):1052– 60. doi: 10.1136/gutjnl-2016-313457 [68] Friedrich M, Pohin M, Jackson MA, Korsunsky I, Bullers SJ, Rue-Albrecht KC, Christoforidou Z, Sathananthan D, Thomas T, Ravindran R, Gagliani N. IL-1-driven stromal–neutrophil interactions define a subset of patients with inflammatory bowel disease that does not respond to therapies. Nature Medicine. 2021;27(11):1970–81. doi: 10.1038/s41591-021-01561-1 [69] Luoma AM, Suo S, Williams HL, Sharova T, Sullivan K, Manos M, Rivero-Gutiérrez B, Sekedat MD, Chan J, Win T, Hosgood HD. Molecular profiling of checkpoint inhibitor–treated melanoma: a single-cell analysis reveals pretreatment markers of response. Cell Reports. 2022;38(13):110320. doi: 10.1016/j.celrep.2022.110320 [70] Der E, Ranabothu S, Suryawanshi H, Akat KM, Clancy R, Morozov P, Kustagi M, Pujato M, Ellis JS, Snyder MP, Sanseviero E. Single cell RNA sequencing to dissect the molecular heterogeneity in lupus nephritis. JCI Insight. 2017;2(12):e93009. doi: 10.1172/jci.insight.93009 [71] Rao DA, Arazi A, Wofsy D, Diamond B. The future of precision medicine in rheumatology: targeting patients and diseases. Nature Reviews Rheumatology. 2021;17(10):621–36. doi: 10.1038/s41584-021-00652-w World Journal of Biology Pharmacy and Health Sciences, 2025, 23(02), 196-218 218 [72] Banchereau R, Hong S, Cantarel B, Baldwin N, Baisch J, Edens M, Cepika AM, Acs P, Turner J, Anguiano E, Vinod P. Personalized immunomonitoring uncovers molecular networks that stratify lupus patients. Cell. 2016;165(3):551–65. doi: 10.1016/j.cell.2016.03.008 [73] Arazi A, Rao DA, Berthier CC, Davidson A, Liu Y, Hoover PJ, Chicoine AT, Eisenhaure TM, Wofsy D, White BP, Gurish MF. The immune cell landscape in kidneys of patients with lupus nephritis. Nature Immunology. 2019;20(7):902– 14. doi: 10.1038/s41590-019-0398-x [74] Kaveri SV, Lacroix-Desmazes S, Mouthon L, Kazatchkine MD. Immunopathogenesis of autoimmune and inflammatory diseases: therapeutic implications. Current Pharmaceutical Design. 2005;11(5):555–73. doi: 10.2174/1381612053382047 [75] Korsunsky I, Fan J, Slowikowski K, Zhang F, Wei K, Baglaenko Y, Brenner M, Loh PR, Raychaudhuri S. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature Methods. 2019;16(12):1289–96. doi: 10.1038/s41592-019-0619-0 [76] Ramilowski JA, Goldberg T, Harshbarger J, Kloppmann E, Lizio M, Satagopam VP, Itoh M, Kawaji H, Carninci P, Rost B, Forrest AR. A draft network of ligand–receptor-mediated multicellular signalling in human. Nature Communications. 2015;6:7866. doi: 10.1038/ncomms8866 [77] Fletcher AL, Lukacs-Kornek V, Reynoso ED, Pinner SE, Bellemare-Pelletier A, Curry MS, Collier AR, Boyd RL, Turley SJ. Lymph node fibroblastic reticular cells directly present peripheral tissue antigen under steady-state and inflammatory conditions. The Journal of Experimental Medicine. 2010;207(4):689–97. doi: 10.1084/jem.20092697 [78] Wen W, Su W, Tang H, Le W, Zhang X, Zheng Y, Liu X, Xie L, Li J, Ye J, Dong L. Immune cell profiling of COVID-19 patients in the recovery stage by single-cell sequencing. Cell Discovery. 2020;6:31. doi: 10.1038/s41421-0200168-9 [79] Croft AP, Campos J, Jansen K, Turner JD, Marshall J, Attar M, Savary L, Wehmeyer C, Naylor AJ, Kemble S, Begum J. Distinct fibroblast subsets drive inflammation and damage in arthritis. Nature. 2019;570(7760):246–51. doi: 10.1038/s41586-019-1263-7 [80] Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine. 2019;25(1):44–56. doi: 10.1038/s41591-018-0300-7 [81] Bianchi DW, Chiu RWK. Sequencing of circulating cell-free DNA during pregnancy. The New England Journal of Medicine. 2018;379(5):464–73. doi: 10.1056/NEJMra1801110 [82] Chen R, Stewart WF, Sun J, Ng K, Yan X. Recurrent neural networks for early detection of heart failure from longitudinal electronic health record data: Implications for temporal modeling in healthcare. Scientific Reports. 2019;9(1):1–10. doi: 10.1038/s41598-019-54849-w [83] Arisdakessian C, Poirion OB, Yunits B, Zhu X, Garmire LX. DeepImpute: an accurate, fast, and scalable deep neural network method to impute single-cell RNA-seq data. Genome Biology. 2019;20(1):211. doi: 10.1186/s13059019-1837-6 [84] Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, Katz R, Himmelfarb J, Bansal N, Lee SI. From local explanations to global understanding with explainable AI for trees. Nature Machine Intelligence. 2020;2(1):56– 67. doi: 10.1038/s42256-019-0138-9 [85] Price WN, Cohen IG. Privacy in the age of medical big data. Nature Medicine. 2019;25(1 [86] Regev A, Teichmann SA, Lander ES, Amit I, Benoist C, Birney E, Bodenmiller B, Campbell PJ, Carninci P, Clatworthy M, et al. The Human Cell Atlas. eLife. 2017;6:e27041. doi: 10.7554/eLife.27041 [87] Bhattacharya S, Dunn P, Thomas CG, Smith B, Schaefer H, Chen J, Hu Z, Zalocusky K, Shankar RD, Shen-Orr SS, et al. ImmPort: disseminating data to the public for the future of immunology. Immunologic Research. 2014;58(2– 3):234–9. doi: 10.1007/s12026-014-8516-1 [88] Singlera Genomics. ImmunoBank Platform Overview. https://www.singlera.com/immunobank [Accessed 202506-15] [89] Foundation Medicine. FoundationOne® Heme Assay Information. https://www.foundationmedicine.com/test/foundation-one-heme [Accessed 2025-06-15] [90] Tempus Labs. Tempus xT + ImmunoINSIGHT Solutions. https://www.tempus.com/oncology/immunoinsight [Accessed 2025-06-15]