Application of spatial omics, multi-omics, and artificial intelligence in inflammatory bowel disease: Current applications and emerging opportunities

Abstract

Inflammatory bowel disease (IBD), encompassing Crohn’s disease and ulcerative colitis, is a complex and heterogeneous disorder driven by dynamic interactions between genetic susceptibility, immune dysregulation, epithelial barrier dysfunction, and the intestinal microbiome. Despite significant therapeutic advances, durable remission remains elusive for many patients, reflecting an incomplete understanding of disease mechanisms within the spatially organized intestinal mucosa. Recent advances in multi-omics and spatial-omics technologies have transformed IBD research by enabling high-resolution molecular profiling while preserving tissue architecture, discovering disease-defining cellular niches and host-microbiome interactions. Simultaneously, artificial intelligence (AI) has emerged as an essential framework for integrating high-dimensional omics datasets, uncovering non-linear relationships, and translating molecular complexity into clinical applications. This review has discussed current applications of multi-omics, spatial omics, and AI-driven analyses that are reshaping disease management. Also, we discussed challenges and emerging opportunities at the intersection of omics and AI, including the development of next-generation non-animal disease models and the discovery of spatial biomarkers for disease progression and cancer risk. Therefore, integration of multi-omics, spatial omics, and AI creates a transformative approach that enhances our understanding and facilitates precision medicine in IBD.

Key Words: Inflammatory bowel disease; Multi-omics; Crohn’s disease; Artificial intelligence; Spatial-omics; Ulcerative colitis; Microbiota

Core Tip: Inflammatory bowel disease (IBD), including Crohn’s disease and ulcerative colitis, arises from complex, spatially organized interactions between host genetics, immune responses, epithelial integrity, and the intestinal microbiota. Emerging multi-omics and spatial-omics technologies now enable high-resolution, tissue-context-preserved profiling of these processes, revealing disease-defining cellular niches and host-microbiome interactions. Artificial intelligence (AI) provides a critical framework for integrating high-dimensional datasets, uncovering non-linear relationships, and translating molecular complexity into clinically actionable insights. The convergence of multi-omics, spatial-omics, and AI holds strong promise for precision medicine in IBD, enabling improved disease stratification, identification of spatial biomarkers for progression and cancer risk, and the development of next-generation, non-animal disease models.

INTRODUCTION

Inflammatory bowel disease (IBD) is a long-term inflammatory condition affecting the digestive system when the body’s natural immune system mistakenly attacks healthy cells as a result of an overactive immune response to standard triggers like food and normal intestinal microbiota. It includes Crohn’s disease (CD), which can affect any part of the gastrointestinal tract, and ulcerative colitis (UC), which affects mainly the colon and rectum part[1-3]. The causes of IBD remain unknown, and several factors such as changes in the function of the immune system, mucosal barrier dysfunction, genetic makeup, environmental triggers, and lifestyle are likely to play a role in its initiation and development of the disease[4,5].

Symptoms and complications associated with IBD comprise extraintestinal anaemia, abdominal discomfort, pain, bleeding, fatigue, diarrhoea, and loss of weight[6-8]. Despite decades of research and the development of multiple therapies, IBD remains incurable, and a significant proportion of patients fail to achieve durable remission[9]. Currently, there is no single treatment for IBD; available treatments range from aminosalicylates, corticosteroids, immunomodulators, biologics, surgery, and nutrition supplements[10-12]. In patients with suspected IBD, diagnosis is established through an integrated clinical assessment, laboratory investigations, cross-sectional radiologic imaging, and endoscopic evaluation with histopathological analysis[13-15].

The intestinal mucosa is a complex tissue featuring a network of epithelial, immune, stromal, vascular, and microbial components arranged into distinct niches[16]. Multi-omics is a general approach in biology that combines data from multiple omics fields (genomics, proteomics, metagenomics, epigenomics, metabolomics, etc.) to get a comprehensive view of biological systems, understand complex diseases, and make discoveries[17]. Recent developments in artificial intelligence (AI) and spatial omics technologies have significantly advanced the study of complex tissues by enabling molecular profiling that retains spatial context (Figure 1). Techniques such as spatial transcriptomics, spatial proteomics, and imaging-based methods facilitate the concurrent mapping of gene expression, protein abundance, and cellular phenotypes within inflamed intestinal tissue[18].

Figure 1 Illustration of a framework that integrates multi-omics, spatial omics, and artificial intelligence for the study of inflammatory bowel disease. Panel (1): Depicts the intestinal mucosa as a spatially organized ecosystem, where disruptions in epithelial integrity, localization of immune cells, and interactions between the host and microbiota lead to disease-specific inflammatory environments in inflammatory bowel disease (IBD); (2) The acquisition of multimodal data merges clinical imaging with spatial omics and traditional multi-omics techniques to capture molecular and cellular changes across various anatomical levels; (3) Artificial intelligence facilitates the integration and spatial examination of these diverse datasets, revealing intricate interactions between the host and microbiome, as well as cellular networks, and identifying predictive features of the disease; and (4) This comprehensive framework advances clinical application by fostering precision medicine approaches for IBD. IBD: Inflammatory bowel disease; AI: Artificial intelligence; MALDI: Matrix assisted laser desorption/ionization.

Rapid expansion of high-dimensional spatial and multi-omics data in IBD has generated large-scale, complex, and multimodal datasets that exceed the capacity of traditional statistical methods, and AI has emerged as an essential tool for extracting biological meaning from those complex datasets. AI is defined as systems (both software and hardware) that display intelligent behavior by learning, analyzing their environment, and taking actions with some degree of independent decision-making to achieve specific goals[19]. AI is changing how clinicians manage IBD by helping in combining different types of data to improve patient stratification and support more personalized care[20]. Integrating AI with spatial-omics and multi-omics will enable the identification of disease-driving variants, dysregulated signaling pathways, metabolic and microbiome-immune interactions[21]. Advancements in IBD research are reshaping traditional clinical and histopathological paradigms. These are driving a transition toward more integrative, data-rich frameworks in which AI-enabled integration can predict disease risk, progression, and disease outcome[22-24]. Historically, disease definition, diagnosis, and management have relied heavily on clinical presentation, endoscopic findings, and single biomarker assessments, often resulting in delayed detection[25-27]. However, emerging approaches employ multi-layered molecular biomarker profiling, spatially resolved tissue analysis, and artificial intelligence-driven data integration to provide a more comprehensive understanding of disease processes[28-30]. These innovations enable earlier and more accurate disease prediction and refined characterization of inflammation at the cellular and subcellular level[31-33]. Shifting from the trial-and-error method of therapy to the development of personalized, mechanism-based treatment strategies[34-36]. In parallel, advances in experimental models and computational simulations are transforming clinical research by enabling more human-relevant and efficient trial designs[37-39]. Together, these shifts highlight a paradigm change from descriptive and reactive care toward predictive, precise, and proactive disease management, as summarized in Table 1.

Table 1 Conceptual transition from conventional clinical and histopathology-based approaches toward next-generation molecular, spatial, and artificial intelligence frameworks in inflammatory bowel disease.

Aspect	Traditional approaches	Emerging approaches
Disease definition	Defined by clinical presentation and histopathological features	Defined by molecular, cellular, and spatial disease states within tissue microenvironments[25-27]
Biomarkers	Detected through single serum or faecal biomarkers (e.g., calprotectin)	Integrated multi-layered molecular signatures across histopathology and omics[28-30]
Inflammation assessment	Analysis of endoscopic and histological scoring systems	Quantitative, spatially resolved inflammation mapped at cellular and subcellular resolution[31-33]
Disease prediction	Often, the disease is detected after the clinical onset	AI-enabled integration of multi-modal molecular, spatial, and clinical data to predict disease risk, progression, and outcomes before or early in the disease course[22-24]
Treatment decisions	Treatment choices depend on empirical evidence, incremental approaches, or a trial-and-error method of increasing therapy	Predictive, mechanism-informed therapy guided by molecular and spatial profiling[34-36]
Clinical trials	Mostly rely on animal models and late-stage human testing	Development and advancement of human-relevant non-animal platforms, including intestinal organoids, gut-on-chip systems, and in silico simulations enabling virtual or hybrid clinical trials in IBD[37-39]

IBD: Inflammatory bowel disease; AI: Artificial intelligence.

This review discusses the current use of spatial omics, multi-omics, and AI in IBD research. Also, translational challenges and significant upcoming opportunities or emerging approaches in IBD that integrate AI, histopathology, and omics technologies to improve understanding and approaches in the clinical management of IBD.

METHODOLOGY

We conducted a comprehensive literature search from PubMed, EMBASE, Web of Science, and Scopus databases to identify relevant studies evaluating the application of spatial omics, multi-omics technologies, and AI in IBD. The search strategy incorporated combinations of medical subject headings and free-text keywords including: (“Inflammatory bowel disease” OR “Crohn’s disease” OR “ulcerative colitis” OR “IBD”) AND (“multi-omics” OR “genomics” OR “transcriptomics” OR “proteomics” OR “metabolomics” OR “metagenomics” OR “epigenomics”) AND (“spatial omics” OR “spatial transcriptomics” OR “spatial proteomics” OR “spatial biology”) AND (“artificial intelligence” OR “machine learning” OR “deep learning” OR “convolutional neural network” OR “neural networks” OR “large language models” OR “AI-driven analysis”). The search covered studies published between January 2006 and January 2026. Greater emphasis was placed on high-impact studies published from 2020 onward, reflecting the rapid technological evolution of spatial omics and AI methodologies. Reference lists of relevant articles and reviews were manually screened to identify additional eligible publications. Inclusion criteria: Original research articles, systematic reviews, and high-impact methodological studies, studies involving human IBD samples, studies applying multi-omics, spatial omics technologies, or AI-based analytical approaches. Exclusion criteria: Conference abstracts without full text, animal-only studies unless directly relevant to translational and studies not specifically addressing IBD, and all articles published in a non-English language.

CURRENT TECHNOLOGIES AND THEIR APPLICATIONS IN IBD

This section will discuss how the recent advances in multi-omics, spatially resolved platforms, and AI have generated diverse high-dimensional datasets that are reshaping IBD research and clinical stratification. Table 2 provides a structured overview of representative studies that summarize key datasets, technologies, analytical frameworks, principal discoveries, and their translational relevance.

Table 2 Summary of representative multi-omics, spatial omics, and artificial intelligence studies in inflammatory bowel disease, highlighting datasets, methodologies, and principal findings.

Ref.	Technology/modality	Sample type	Cohort size	Analytical approach	Key findings	Clinical/translational relevance
Uniken Venema et al[43]	Epigenomics (DNA methylation)	Blood	240 IBD, 190 controls	Epigenome-wide methylation profiling	Differential methylation at VMP1, ITGB2	Links genetic risk loci to environmental regulation
Chan et al[65]	Proteomics (LC-MS/MS)	Colonic biopsies	10 UC, 10 controls	Gel-free quantitative proteomics	High level of neutrophil-related proteins (calprotectin, lactotransferrin)	Tissue-level inflammatory biomarkers
Quiroga-Centeno et al[84]	Metabolomics (HPLC)	Serum	> 530 IBD	Targeted metabolite profiling	Low levels of tryptophan correlate with disease activity	Potential metabolic biomarker
Aichler and Walch[103]	Spatial transcriptomics, scRNA-seq	Intestinal tissue	Human IBD biopsies	Single-cell, spatial mapping	IDA macrophages, NRG1-dependent epithelial regeneration	Defines pathogenic immune niches
Lins Neto et al[97]	Imaging mass cytometry	Intestinal tissue	Human IBD	Multiplex spatial proteomics	Disease-specific immune-cell neighborhoods	Spatial immune architecture as a biomarker
Mund et al[104]	10 × Visium spatial transcriptomics	Postoperative CD biopsies	43 biopsies	Spatial pathway analysis	TNF, NF-κB, and JAK-STAT are conserved across sites	Shared inflammatory programs
Garrido-Trigo et al[113]	AI (CNN model)	Endoscopic images	Large UC dataset	Deep learning severity scoring	Continuous inflammation scoring (UCEGS)	Objective endoscopic grading
Nair et al[161]	Blood transcriptomics	Blood	IBD patients	Machine learning classifier	Prognostic biomarker for disease course	Patient stratification
Kim et al[132]	Spatial multi-omics (MicroCart)	Murine colitis	Experimental	Integrated host-microbiome mapping	Coordinated immune epithelial microbial niches	Spatial host-microbiome cartography

LC-MS/MS: Liquid chromatography-mass spectrometry/mass spectrometry; HPLC: High performance liquid chromatography; scRNA-seq: Single-cell RNA-sequencing; CNN: Convolutional neural network; AI: Artificial intelligence; MicroCart: Microbiome cartography; CD: Crohn’s disease; IBD: Inflammatory bowel disease; UC: Ulcerative colitis; VMP1: Vacuole membrane protein 1; ITGB2: Integrin subunit β2; IDA: Inflammation-dependent activated macrophages; TNF: Tumor necrosis factor; NF-κB: Nuclear factor kappa-B; JAK-STAT: Janus tyrosine kinase/signal transducer and activator of transcription; UCEGS: Ulcerative colitis endoscopic index of severity.

MULTI-OMICS TECHNOLOGIES

IBD is suspected to arise from the convergence of genetic susceptibility, environmental exposures, immune dysregulation, epithelial dysfunction, and microbial imbalance. Each of these dimensions has been cross-examined using individual omics technologies, generating extensive datasets that capture specific aspects of IBD[40]. A growing number of multi-omics platforms have been widely adopted in IBD research, contributing to extensive advances in understanding disease mechanisms and heterogeneity. The following are some of the multi-omics technologies applied in IBD research.

Genomics

These approaches utilise high-throughput technologies to identify and characterize genetic variants and molecular determinants that influence disease susceptibility, progression, and therapeutic response in IBD. Through genome-wide association studies (GWAS) and sequencing-based strategies, these investigations have expanded our understanding of inherited risk architecture and the molecular pathways underlying disease heterogeneity[41]. The first wave of GWAS on IBD provided a major historical pivot from candidate genes to unbiased discovery, identified interleukin (IL)-23R as a key susceptibility locus, and helped establish the IL-23/T helper (Th) 17 axis as a central immune pathway in IBD pathogenesis[42].

GWAS have outlined a broad genetic susceptibility architecture underlying IBD. Among the numerous identified risk loci, some subset of important causal genetic variants, candidate genes such as ATG16 L1, IRGM, NOD2, PRDM1, CARD9, RNF186, and IL23R have undergone detailed functional validation and have been linked to disease-relevant biological pathways, including host microbial sensing, autophagy, epithelial barrier integrity, and immune regulation[43].

Variants in NOD2 remain among the most well-established genetic risk factors for CD, while exome sequencing has uncovered additional coding variants in genes involved in immune regulation, including PRDM1, NDP52, IL17REL, and CSF2RB. Identification of protective genetic variants provides a complementary strategy for drug discovery. For example, loss-of-function variants in RNF186 are associated with reduced risk of UC, highlighting this pathway as a potential target for therapeutic modulation in IBD[44].

GWAS have been instrumental in identifying genetic loci that confer susceptibility to IBD, including key immune-regulatory genes such as IL23R, IL12B, JAK2, and STAT3[45]. These discoveries have converged on central inflammatory signaling networks, mainly those governing Th17 cell differentiation and the IL-12/IL-23 cytokine axis, providing critical insight into the immune pathways driving IBD pathogenesis[45]. It had informed on disease mechanisms by the identification of genetic variants and candidate gene networks that affect host-microbe interactions, for example, microbe sensing within the NOD2 pathway. Also, it has helped in the identification of autophagy as a pathogenic mechanism in IBD, highlighting the key role of inflammatory signaling pathways such as IL-23-driven Th cell response[44].

Genomics analyses have reported several genetic risks for IBD, as characterized by enrichment of immune-related genes that are preferentially expressed in the spleen, terminal ileum, and congregate inflammatory pathways such as Th17 differentiation, cytokine signaling, and tumor necrosis factor (TNF)-mediated responses, highlighting shared mechanisms across immune-mediated diseases[46]. Recently, the application of genetic testing and DNA sequencing helped researchers in identifying new single-nucleotide polymorphisms associated with IBD, like those near IL23R (rs11209026) and IFIH1 (rs1990760)[46].

A whole-genome sequencing analysis study was conducted on more than 4200 patients with UC and CD, combined with more than 3600 population controls. The results of the study identified a low-frequency missense variant in ADCY7 that was associated with a substantially increased risk of UC. This finding highlights the contribution of rare coding variation to IBD susceptibility and underscores the value of genome-wide sequencing approaches for uncovering functionally relevant disease risk alleles[47].

Recent advances in IBD diagnostics have incorporated high-throughput genomic technologies combined with machine learning approaches to improve early detection and biomarker discovery. Gene expression profiles derived from colonic biopsy specimens of 172 patients with IBD and 22 healthy controls were analyzed using the publicly available GSE75214 microarray dataset. Comparative transcriptomic analysis, machine learning algorithms, and feature selection strategies identified six differentially expressed genes, VWF, IL1RL1, DENND2B, MMP14, NAAA, and PANK1, that demonstrated strong diagnostic potential[48]. These biomarkers were proposed as candidate molecular signatures that may facilitate earlier detection and more accurate classification of IBD[48].

Integrative analyses of four IBD datasets from the Gene Expression Omnibus database have identified a subset of genes with strong disease-associated signatures, including IRF1, GBP5, and PARP9, which consistently exhibit pro-inflammatory activity in IBD. These genes are implicated in disease pathogenesis and represent promising candidates as both diagnostic biomarkers and potential therapeutic targets, highlighting the value of data-driven genomic analyses in advancing precision medicine for IBD[49].

Epigenomics

IBD may arise from heritable changes in gene function that are not caused by alterations in the DNA sequence but from regulatory networks, such as DNA methylation, histone modifications, and non-coding RNA expression[50]. Intestinal development is profoundly influenced by early-life interactions with dietary components, commensal microorganisms, and antigenic stimuli that drive mucosal immune maturation and immune tolerance[51]. Disruption of these tightly regulated processes by factors such as enteric infections, toxic exposures, or pharmacological agents can induce lasting epigenetic alterations that perturb host microbiome homeostasis and increase long-term risk of intestinal inflammation[45,52].

An epigenome-wide DNA methylation analysis study of 240 newly diagnosed patients with IBD, compared with 190 healthy controls, identified disease-associated differentially methylated regions at vacuole membrane protein 1 and integrin subunit β2, as well as differentially methylated positions at RPS6KA2, a member of the serine/threonine ribosomal kinase family[53]. These epigenetic changes overlapped with known genetic risk loci, highlighting DNA methylation as a key link between inherited susceptibility and environmental exposures in IBD[53].

DNA methylation within promoter regions or other critical regulatory elements can repress transcriptional activity, effectively reducing or silencing gene expression[54]. This epigenetic mechanism plays a vital role in regulating genes involved in immune signaling and epithelial homeostasis. In IBD, aberrant methylation at the CLDN2 locus has been associated with impaired intestinal barrier integrity. CLDN2 encodes claudin-2, a tight junction protein essential for regulating paracellular permeability, and increased methylation at this site has been associated with reduced gene expression. The resulting disruption of tight junction function weakens mucosal barrier defences, thereby heightening vulnerability to inflammation-induced epithelial injury[55].

Emerging evidence indicates that epigenetic regulation, mostly DNA methylation, contributes to interindividual variability in therapeutic responses across inflammatory disorders. Methylation changes within regulatory regions of the TNF have been shown to modulate expression of TNF-α, a key cytokine driving inflammatory pathology. Dysregulation of methylation at these loci can shift TNF-α production in either direction, thereby influencing clinical responsiveness to anti-TNF biologics that are widely used in the treatment of rheumatoid arthritis and IBD[55].

Modulation of H3K27 trimethylation has been shown to influence Th17 cell differentiation in UC[56]. Pharmacological inhibition of the H3K27 methyltransferase EZH2 promotes the expansion of anti-inflammatory myeloid-derived suppressor cells, highlighting a potential therapeutic avenue for immune modulation in IBD[56,57]. In paediatric IBD, it was found that sustained enrichment of histone H3-lysine 4 trimethylation (H3K4me3) mark at inflammation-associated gene loci correlates with disease severity, emphasizing the role of epigenetic priming early in the disease course[58]. Additional histone-modifying enzymes, including the H4K20 methyltransferase SETD8, have been implicated in regulating inflammatory gene expression (regulating p62 expression) and colitis progression through locus-specific chromatin remodeling[59]. These findings suggested histone modifications as dynamic and therapeutically targetable regulators of immune and epithelial dysfunction in IBD.

Non-coding RNAs are involved in the molecular pathophysiology of IBD through the regulation of gene expression and molecular pathways[60,61]. For example, in CD, elevated expression of specific long non-coding RNAs (lncRNAs) has been detected in patients with active disease, including DQ786243, which is significantly increased in peripheral blood cells compared with healthy individuals. In addition, lncRNAs implicated in CD include NRON, a negative regulator of nuclear factor of activated T cells (NFAT) signaling that functions within RNA-protein complexes to restrict NFAT nuclear translocation and is functionally linked to immune regulatory pathways involving susceptibility genes such as LRRK2[62]. In UC, distinct lncRNA signatures have also been identified, with BC012900 markedly upregulated in inflamed colonic tissue and inducible by cytokine and microbial stimuli through key innate immune pathways, including Toll-like receptor and NOD2 signaling[60].

LncRNAs have emerged as potential biomarkers and predictors of therapeutic response in IBD[63]. A study analyzed the role of non-coding RNA in IBD and found that alterations in the expression of the lncRNA such as ANRIL were associated with clinical response to infliximab therapy. Increased mucosal expression of ANRIL has been observed in patients with CD who respond to treatment, whereas expression levels remain unchanged in non-responders[64].

Transcriptomics

Profiling of gene expressions can identify RNA transcript changes associated with IBD[65]. Currently, transcriptomic profiling at single-cell resolution has dramatically expanded understanding of cellular diversity in IBD[66]. Various analytical techniques have been used in IBD, including polymerase chain reaction, microarrays, bulk RNA sequencing, and single-cell RNA sequencing, which is a cutting-edge technology that enables the exploration of individual cells genetic activity within tissues[67].

Transcriptomic profiling has been widely used to dissect the heterogeneous molecular and cellular mechanisms underlying UC, enabling characterization of immune cell composition, dysregulated pathways, and differentially expressed genes across patient cohorts, as well as their associations with therapeutic outcomes[68]. UC tissues commonly display an imbalance in macrophage polarization, characterized by an expansion of pro-inflammatory M1 macrophages alongside a relative reduction in anti-inflammatory M2 populations. Through their functional plasticity, macrophages can amplify inflammatory responses by reinforcing reciprocal activation loops, enhancing dendritic cell activation, and coordinating with adaptive immune cells such as T and B lymphocytes, thereby sustaining mucosal inflammation[69].

Integrated transcriptomic analyses of peripheral blood samples from patients with IBD have identified FEZ1 as a candidate circulating biomarker associated with disease status. Although current evidence suggests that the involvement of FEZ1 in IBD may be linked to autophagy-related pathways, its precise immunological functions and mechanistic contribution to disease pathogenesis remain to be fully elucidated through further investigation[70]. A multi-tissue transcriptome-wide association study of IBD samples identified TPRKB and RNF149 as new IBD susceptibility genes. In the same survey, it was also found that TPRKB and RNF149 have a close relationship with TP53RK and LAMP1 respectively[71].

Integrative analyses of public microarray and RNA-sequencing (RNA-seq) datasets from 36 patients with IBD and 30 healthy controls, revealed distinct immune signatures in IBD, characterized by increased M0 macrophages, regulatory T cells, and naive cluster of differentiation 4 (CD4)⁺ T cells, alongside reduced memory B cells and activated natural killer cells[72]. Network-based analyses identified NDUFB2 as a central downregulated hub gene, linking IBD-associated transcriptional changes to mitochondrial complex I dysfunction and impaired oxidative phosphorylation, highlighting systemic metabolic alterations as potential diagnostic features[72].

Proteomics

Current proteomic technologies have enabled comprehensive characterization of protein abundance in tissues, blood, or stool to determine their structure and functional implications in the initiation and development of disease. These methods provide deeper insights into all proteins expressed[73], molecular mechanisms underlying IBD by enabling direct characterization of protein expression within intestinal tissues[74].

A study analyzed intestinal biopsies using gel-free quantitative proteomics on colonic mucosal biopsies from 10 UC patients and 10 controls. More than 5700 proteins were quantified, and found that tissue levels of neutrophil-related proteins such as lactotransferrin and calprotectin show strong correlations with histologically assessed tissue inflammation[75]. These findings indicate that activation of innate immune pathways, mostly neutrophil-driven antimicrobial responses, is a prominent feature of UC pathogenesis[75].

A broad range of circulating biomarkers has been investigated for their utility in IBD, encompassing acute-phase reactants, hematologic parameters, complement components, and inflammatory cytokines. Among these, C-reactive protein has emerged as the most extensively used serum biomarker in clinical practice, owing to its widespread availability and established role in monitoring inflammatory activity in patients with IBD[76].

Faecal biomarkers offer enhanced specificity for inflammatory processes localized to the intestinal mucosa and have become integral to non-invasive assessment of IBD[77]. While markers reflecting epithelial barrier disruption, such as α-1-antitrypsin or faecal occult blood, provide indirect evidence of mucosal injury, biomarkers linked to leukocyte activation closely reflect pathogenic inflammatory activity[77]. Proteins released during leukocyte degranulation, including lactoferrin, myeloperoxidase, elastase, matrix metalloproteinase-9, and neopterin, are consistently elevated in the stool of patients with active IBD. In addition, damage-associated molecular pattern proteins such as calprotectin are highly stable in faeces and robustly correlate with intestinal inflammation, as do faecal M2-pyruvate kinase and chitinase-3-like-1. Among available assays, calprotectin and lactoferrin remain the most extensively validated and widely implemented in routine clinical practice[77].

Microbiomics

Gut microbiota plays a central role in disease initiation and development of IBD[78]. IBD is widely thought to arise, in part, from intestinal dysbiosis characterized by an imbalance between pro-inflammatory microbial populations and beneficial anaerobic commensals[79]. Early culture-independent, 16S rRNA gene-based profiling studies provided some of the first high-confidence evidence that IBD is associated with community-level dysbiosis, including reduced abundance of key obligate anaerobes (especially Firmicutes and Bacteroidetes) and expansion of pathobiont-enriched configurations in subsets of CD and UC[80]. These foundational observations helped frame IBD as a disorder of host microbiota ecological imbalance and immune miscalibration rather than infection with a single classical pathogen[80].

Alterations in gut microbiota composition are recognized as a key contributor to the development and progression of IBD[81]. In healthy individuals, the intestinal ecosystem comprises a diverse array of microbial species that support metabolic and barrier functions. In contrast, IBD is characterized by a depletion of beneficial butyrate-producing bacteria alongside an expansion of microbial populations with enhanced mucin-degrading capacity[73,82].

High-resolution profiling of the gut microbiome using 16S rRNA gene sequencing and metagenomic approaches from 18 patients with IBD revealed disease-specific microbial signatures compared with healthy individuals. CD was characterized by a marked enrichment of Proteobacteria members of the Escherichia and Shigella genus, whereas UC was associated with increased abundance of Enterococcus[83]. In contrast, several core commensal taxa, including members of the Christensenellaceae R-7 group, Fusicatenibacter, and Holdemanella, were predominantly observed in healthy controls. Moreover, integrative analyses uncovered significant associations between microbial composition and virulence-related pathways, supporting a potential role for oxidative stress-driven host-microbe interactions in shaping disease-specific microbial dysbiosis in IBD[83].

High-throughput sequencing approaches and dysbiosis mapping tools enable detailed characterization of microbial community shifts associated with IBD and related functional bowel disorders[84]. Faecal microbiota signatures have been shown to distinguish active from quiescent CD, underscoring their potential as non-invasive indicators of disease activity. Specific microbial markers, including molecular signatures derived from Campylobacter concisus, have been associated with active disease states. In addition, accumulating evidence supports the utility of faecal microbiota profiles for monitoring therapeutic response, with increased abundance of faecal bacterium species linked to favorable outcomes following ustekinumab treatment in patients with anti-TNF refractory CD[85].

Therapeutic interventions in IBD can induce measurable alterations in both the composition and functional capacity of the gut microbiome. Treatment with biologic agents, including anti-TNF, anti-integrin, and anti-IL therapies, has been associated with increased microbial diversity and partial restoration of intestinal eubiosis in patients with CD[86]. These shifts are characterized by enrichment of short-chain fatty acid-producing taxa, such as members of the Firmicutes and Bacteroidetes phyla, alongside a reduction in Proteobacteria. Therefore, treatment-associated microbiome remodeling supports the potential utility of microbial signatures as biomarkers for predicting therapeutic response in IBD[86].

Metabolomics

Biologically produced small molecules (metabolites) can be sufficient to discriminate and diagnose complex diseases[87]. Metabolomics has emerged as a powerful approach for defining metabolic signatures that distinguish patients with IBD from healthy individuals[88-90]. Due to its high sensitivity, analytical precision, and broad molecular coverage, metabolomic analyses have identified disease-associated metabolites across multiple biospecimens, including serum, faeces, urine, and intestinal biopsies. By capturing downstream biochemicals, metabolomics represents a powerful platform for the discovery of biomarkers relevant to IBD diagnosis, prognosis, and therapeutic stratification[91].

Disruption of the gut microbiota alters the production of key metabolites that contribute to IBD pathogenesis[92]. Patients with IBD commonly exhibit reduced levels of short-chain fatty acids, reflecting the depletion of butyrate-producing taxa, including Faecalibacterium prausnitzii and members of Clostridium clusters IV, XIVa, and XVIII. Diminished short-chain fatty acids availability impairs regulatory T cell differentiation, epithelial cell renewal, and overall intestinal homeostasis. In parallel, expansion of sulfate-reducing bacteria, including Desulfovibrio species, increases hydrogen sulfide production, which compromises epithelial integrity and promotes mucosal inflammation[93].

Large-scale metabolic profiling of serum samples from more than 530 patients with IBD, using high-performance liquid chromatography, has discovered a strong inverse association between circulating tryptophan levels and disease activity. Elevated concentrations of tryptophan catabolites, such as quinolinic acid, in patients with active disease indicate enhanced activation of tryptophan degradation pathways[94]. Dysregulated tryptophan metabolism may contribute to disease pathogenesis or exacerbate inflammation, highlighting intestinal and systemic tryptophan pathways as potential therapeutic targets. However, interventional studies are required to determine whether modulation of tryptophan metabolism can alter disease severity[94].

Integrated analyses of faecal microbiome composition and metabolite profiles have revealed characteristic shifts in microbial communities, including depletion of health-associated phyla such as Bacteroidetes and Firmicutes, alongside expansion of disease-associated taxa, especially Proteobacteria and Actinobacteria, with enrichment of species such as Escherichia coli and Klebsiella pneumoniae[95]. High-resolution metabolomic profiling has identified a subset of metabolites with strong discriminatory power between IBD and control samples, including 6,7,4’-trihydroxyisoflavone and thyroxine 4’-o-beta-d-glucuronide, normorphine, salvinorin A, and trichostachine, highlighting candidate diagnostic biomarkers. These metabolic alterations converge on key pathways involved in primary bile acid biosynthesis, vitamin digestion, absorption, and carbohydrate metabolism, underlining the profound metabolic reprogramming associated with intestinal inflammation and dysbiosis in IBD[95].

Metabolomics has enabled non-invasive strategies for distinguishing IBD subtypes based on disease-specific metabolic signatures[96]. Faecal metabolomic profiling using high-resolution proton nuclear magnetic resonance spectroscopy, combined with multivariate analytical approaches, has demonstrated the capacity to differentiate CD from UC. Distinct metabolite patterns were observed across disease subtypes, with enrichment of metabolites such as lactate, succinate, alanine, and tyrosine in CD, and leucine, alanine, and tyrosine in UC[97].

SPATIAL OMICS IN IBD

The recognition of IBD as a spatially organized systems disease has catalyzed the adoption of spatial omics technologies that preserve tissue architecture while enabling high-dimensional molecular profiling. Combining spatial omics with microbiome approaches offers a unique opportunity to interrogate IBD[98]. In recent years, different spatial omics technologies have been increasingly applied to study IBD.

Spatial proteomics

Recent advances in spatial proteomic technologies have expanded our ability to characterize the location, quantity, and interactions of proteins within cells and tissues. By preserving spatial context while capturing their complex spatial organization and dynamics, which helps to understand cellular function and disease-associated alterations[99]. Spatial proteomics enables high-resolution mapping of protein expression and distribution across intestinal epithelial, immune, and intrinsic muscular layers, while also capturing spatial heterogeneity in host-microbiota interactions within the gut microenvironment, providing deep insights into how their distribution affects biological processes and IBD state[100].

Various spatial proteomics methods have been applied to study IBD, including imaging-based spatial proteomics, which relies on antibody-mediated detection of proteins directly within intact tissue sections, thereby preserving cellular morphology and spatial relationships, e.g., immunohistochemistry and immunofluorescence[101]. Affinity-based spatial proteomics quantifies protein abundance within defined regions of interest using an antibody-linked molecular barcode, e.g., digital spatial profiling[102]. Mass spectrometry-based spatial proteomics, which relies on label-free mass spectrometry to measure proteins directly from tissue sections, e.g., matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry[103]. Deep visual proteomics (DVP), which combines artificial-intelligence-driven image analysis of cellular phenotypes with automated single-cell or single-nucleus laser microdissection and ultra-high-sensitivity mass spectrometry[104]. Mass spectrometry-based platforms such as MALDI-time of flight and surface-enhanced laser desorption/ionization-time of flight enable sensitive, high-throughput quantification of protein abundance from diverse biological specimens and facilitate comparative profiling between patients with IBD and healthy individuals[105].

High-resolution spatial proteomic approaches, including DVP, have enabled comprehensive profiling of intestinal epithelial cell states within the human colon. By integrating AI-guided image analysis with ultra-sensitive mass spectrometry, DVP has generated detailed proteomic maps of colonic mucosa and facilitated the validation of intestinal organoid models that present in vivo epithelial biology relevant to IBD[106]. Especially, the depth and sensitivity of this approach allow reliable quantification of low-abundance regulatory proteins, including stem cell associated markers such as SOX9 and LGR5, which are critical for epithelial renewal and regeneration. These capabilities position DVP as a powerful platform for dissecting epithelial heterogeneity and advancing translational studies of epithelial dysfunction and therapeutic response in IBD[106].

High-dimensional single-cell spatial analyses have revealed that the intestinal epithelium in IBD exhibits increased proliferative activity and elevated expression of antigen-presenting molecules, features that correlate with histologic inflammation severity[107]. Spatial neighbourhood analyses further uncovered disease-specific immune interaction networks, including enrichment of regulatory T cell proximity to macrophages, CD4⁺ T cells, and plasma cells across IBD, with distinct macrophage subsets preferentially interacting with regulatory T cells in CD compared with UC[107].

Spatial transcriptomics

Recent advances in spatially resolved transcriptomic technologies have significantly expanded our ability to map RNA abundance directly within intact tissue. By preserving positional information while capturing transcriptional activity, these approaches enable gene expression programs to be interpreted within their native cellular and tissue context[108]. Current spatial transcriptomics methods applicable to IBD include imaging-based approaches, such as in situ hybridization and sequencing-based methods, e.g., single-molecule fluorescence in situ hybridization and multiplexed error-robust fluorescence in situ hybridization (FISH). Capture-based spatial transcriptomics, which uses spatially barcoded arrays or beads to capture transcripts from defined tissue locations, enabling transcriptome-wide profiling at near single-cell resolution, e.g., Visium and Slide-seq. Sequencing-based in situ spatial transcriptomics, which combines iterative hybridization and sequencing cycles to achieve high multiplexing and spatial resolution, e.g., seqFISH and MERFISH[109].

A recent study developed a high-throughput multiplex immunofluorescence platform based on opal chemistry to characterize T cell populations in murine colonic tissue, with particular emphasis on activation and proliferation markers to capture functional immune states within the intestinal microenvironment. AI-assisted image analysis workflow was incorporated to enable rapid, reproducible, and quantitative evaluation of spatially resolved cellular phenotypes. This combined experimental and computational approach has demonstrated flexibility across preclinical rodent models, suggesting its broader application for investigating immune dynamics and evaluating T cell-centric therapeutic strategies in translational IBD research[110].

Analyses of terminal ileal biopsies from paediatric patients with ileal CD discovered induction of specific lncRNAs, including LINC02739, in regions of bacterial penetration, implicating these transcripts in host-microbe crosstalk and disease pathogenesis[111]. Comparative spatial analyses further demonstrated that epithelial cells exposed to beneficial microbial communities exhibited increased expression of protective and regulatory genes, including REG1A and TNFRSF6B. In contrast, pathogenic microbial niches were associated with inflammatory signatures. This approach enabled the identification of previously uncharacterized beneficial bacterial species within the Bacillota (Firmicutes) phylum, including Sellimonas intestinalis, Turicimonas muris, and Dysosmobacter welbionis, which were phylogenetically similar to known health-associated microbes[111]. Pathogenic microbial populations were enriched for members of the Pseudomonadota (Proteobacteria) phylum, consistent with prior observations linking Proteobacteria expansion to IBD[112]. These findings illustrate how spatial transcriptomics provides a new framework for dissecting localized host-microbiome interactions in IBD.

Spatial transcriptomics and single-cell RNA sequencing identified diverse myeloid populations in IBD, including resident macrophage states (M0 and M2), inflammation-dependent activated macrophages (IDA), infiltrating monocytes, and distinct neutrophil subsets (N2/CXCR4⁺ neutrophils), highlighting marked spatial and inter-patient heterogeneity within the inflamed mucosa. Inflammatory macrophages displayed alternative activation signatures rather than classical M1 profiles, characterized by expression of epidermal growth factor receptor (EGFR) ligands (NRG1, HBEGF), C-type lectin receptors (CLEC10A, ASGR1), and responsiveness to endogenous signals such as serotonin, potentially derived from enterochromaffin cells[113]. Spatial localization revealed NRG1 with high IDA macrophages enriched near the epithelial surface, where they may promote epithelial regeneration via EGFR signaling and induction of OLFM4⁺ regenerative epithelial states, whereas NRG1 with low IDA macrophages localized to granulomas and submucosal regions in CD. Additionally, spatial interactions between M2 or IDA macrophages and inflammatory fibroblasts expressing IL-11, IL-24, IL-8, IL-6, transforming growth factor-β1, CSF2 (granulocyte-macrophage colony stimulating factor), and matrix metalloproteinases suggested a pathogenic macrophage-fibroblast crosstalk driven by mediators such as oncostatin M, TNF, IL-6, and EGFR signaling that may contribute to chronic inflammation, fibrosis, and tissue remodeling in IBD[113].

Analysis of 43 biopsies from the colon, neo-terminal ileum, and anastomosis using the 10 × Genomics Visium platform demonstrated that inflamed regions are enriched for T cells, B cells, and myeloid populations[114]. At the same time, non-inflamed areas are dominated by epithelial cells. Inflamed niches showed coordinated activation of TNF-α, nuclear factor kappa-B, Janus tyrosine kinase/signal transducer and activator of transcription (JAK-STAT), EGFR, and mitogen-activated protein kinase (MAPK) signaling pathways, highlighting shared molecular drivers of postoperative inflammation independent of tissue location[114].

Spatial proteogenomics

Spatial proteogenomics has recently emerged as a powerful hybrid strategy in tissue biology, combining genomic variation, transcriptional activity, and protein-level signaling within their native spatial context. Rather than examining these layers independently, this approach connects gene regulation to functional protein states directly in situ, enabling a more coherent understanding of disease-associated cellular interactions and signaling networks[115]. Different spatial-omics technologies, such as imaging mass cytometry, enable highly multiplexed, single-cell resolution mapping of protein expression within intact tissue architecture[107].

The integration of single-cell RNA sequencing with spatial molecular imaging enables the construction of high-resolution maps of both healthy and inflamed colonic tissue[113]. The most advanced applications of spatial proteomics arise from its combination with spatial transcriptomics, allowing direct assessment of whether transcriptional programs are translated into functional protein outputs. Such multi-modal spatial integration has discovered discrete cytokine-enriched microenvironments that are not fully captured by RNA-based analyses alone, thereby providing a more accurate and functionally informed representation of the inflammatory landscape in IBD[113]. Integrative single-cell and spatial multi-omics analyses have identified mononuclear phagocytes as a key cellular population modulated by anti-integrin therapy in UC, alongside coordinated alterations in stromal and epithelial cell states[116].

A spatial-omics study of 3 patients with long-standing colitis, compared with healthy controls, discovered spatially restricted activation of pathways such as IL-17 signaling and JAK-STAT signaling in both epithelial and immune compartments, highlighting that chronic, classically defined inflammation, rather than global inflammatory burden, may underlie dysplasia risk[98]. These findings support the concept that spatially defined inflammatory signatures could serve as early biomarkers for cancer risk stratification and inform surveillance and prevention strategies in long-standing IBD[98,117].

ARTIFICIAL INTELLIGENCE IN IBD

AI has been incorporated in the field of medicine and is utilized across diagnosis, prevention, prognosis, and treatment prediction. The following section discusses the incorporation of AI in IBD.

Integration of multi-omics, radio-pathomics, and spatial data in IBD research

One of AI’s transformative roles in IBD lies in the integration of heterogeneous data to improve patient stratification and enable personalized care. By combining multi-omics, radiologic imaging, and longitudinal clinical metadata, AI provides a systems-level framework for disease characterization[118], effectively identifying non-linear associations and latent structures within complex datasets. Ultimately, enabling the discovery of disease-relevant patterns that may remain inaccessible to conventional analyses[30]. Integrating AI with spatial-omics and multi-omics addresses key challenges in gastrointestinal research, such as the identification of disease-driving variants, dysregulated signaling pathways, metabolic and microbiome-immune interactions[21]. AI is expected to reshape the diagnosis, classification, and longitudinal monitoring of IBD, addressing key limitations inherent to current clinical workflows[119]. This integrative approach positions AI as a central platform for advancing precision medicine in IBD. One prominent example of AI-driven integration is the analysis of the SPARC IBD cohort, where AI models were used to harmonize genomics, gut biopsy transcriptomics, and blood plasma proteomics from hundreds of patients[120]. This computational framework successfully identified two distinct subpopulations of CD characterized by unique inflammatory profiles that are invisible to traditional clinical assessments[120].

Standardization of endoscopic image, histology assessment, and disease activity scoring

AI-driven image analysis has significantly refined the evaluation of endoscopic disease activity in IBD. In addition, AI tools can assess intestinal barrier integrity and healing, providing new insights into deep mucosal recovery, going beyond traditional clinical and endoscopic measures[20]. Convolutional neural network (CNN)-based systems can distinguish CD from UC, quantify inflammatory severity, detect histologic remission, and identify gastrointestinal bleeding sources[121]. CNNs were trained on comparative pairs of images to quantify inflammation on a continuous scale, the results suggested improved sensitivity for detecting mucosal healing and treatment response compared with standard scoring systems[122]. Across multiple studies, AI systems have achieved performance levels comparable to experienced IBD endoscopists, supporting objective and reproducible disease assessment[123]. AI-based analysis of pathological slides automates the quantification of inflammatory infiltrates, such as neutrophils and eosinophils, in whole-slide biopsies[124]. To enhance precision endoscopy and histology, AI can differentiate various cell types pivotal to IBD pathogenesis with accuracy levels comparable to experienced pathologists, allowing for a standardized identification of histological remission and a more objective grading of chronic active colitis assessment[123]. Currently, one of the most impactful clinical applications of AI in IBD is image-based histopathological evaluation[125]. CNN and related deep learning models have demonstrated high accuracy in detecting and grading endoscopic inflammation, distinguishing UC from CD, and identifying mucosal healing more sensitively than conventional scoring systems[126]. Advanced AI systems have also been applied to capsule endoscopy for lesion detection in CD, improving diagnostic sensitivity and reducing observer variability in complex imaging tasks[127]. To improve endoscopic diagnosis and outcome assessment in IBD, P0245, a prospective FAIR-compliant video database for an AI model was created. This prospective video databank was established at Montreal University Hospital centre with the enrolment of IBD patients undergoing endoscopy. This multimodal video and image database for AI training will further enhance these capabilities and support precision endoscopy research. As of December 2025, the databank includes 8512 endoscopic procedures, including 1207 colonoscopies in IBD patients[128].

Prediction of therapeutic response and disease course

AI-driven approaches are increasingly used to predict therapeutic response and identify dysplasia, supporting real-time clinical decision-making. By applying in silico approaches to integrate multimodal datasets, including genomic variation and digital histopathology, AI can forecast which patients are likely to achieve mucosal healing before a biologic therapy is initiated[129]. Such in silico approaches facilitate virtual experimentation and hypothesis generation, accelerating disease detection and informing treatment optimization. Furthermore, recent benchmarks show that large language models (LLMs) can align complex therapeutic decisions with international standards, such as the European Crohn’s and Colitis Organization (ECCO) guidelines[130]. AI systems facilitate more consistent and evidence-based management across healthcare settings. A recent multi-centre study of 110 IBD cases from two Mexican institutions evaluated six LLMs. For each case, AI-generated treatment recommendations were compared with those from gastroenterologists and the ECCO guidelines; the overall agreement with ECCO guidelines was high[130]. These results indicate a shift from using AI solely as an analytical tool to adopting it as a structured clinical co-pilot in IBD management. A recent multi-centre study involved more than 740 patients with CD. An eXtreme Gradient Boosting-based model was developed that accurately distinguished long-term responders from non-responders to infliximab[131]. It was suggested that such predictive models can help clinicians identify patients most likely to benefit from specific biologic therapies and tailor treatment plans accordingly[131]. Moreover, recurrent neural network-based models were developed on 149 IBD patients to estimate the concentration of infliximab[132]. The model achieved high area under the curve values, further underscoring the potential for AI-guided precision therapy in IBD[132].

AI for clinical workflow and data structuring

Advances in AI are improving efficiency in IBD care by automating documentation. In parallel, advances in LLMs are improving efficiency in IBD care through restructuring routine clinical workflows by automating documents, literature retrieval, and electronic health record review[133]. Recent work has evaluated LLMs for transforming unstructured IBD histopathology and radiology reports into standardized, ontology-linked representations. These models capture key features such as anatomical location and diagnostic qualifiers with high accuracy, allowing clinicians to securely and reliably structure electronic health record data. This technical path accelerates large-scale research and enables downstream integrative analyses by turning routine clinical text into actionable research data[134]. LLMs are increasingly used to streamline documentation by extracting key imaging features from free-text magnetic resonance enterography (MRE) reports and automatically calculating validated activity indices such as the simplified magnetic resonance index of activity (sMARIA). In a retrospective study of 117 anonymized MRE reports from patients with confirmed CD, ChatGPT (GPT-4o) and DeepSeek (R1) were evaluated for feature extraction and score computation[135]. Performance was assessed using sensitivity, specificity, precision, F1-score, and error analysis. Overall, the findings indicated that LLMs can extract structured data and compute sMARIA scores with near-expert accuracy directly from routine reports[135]. This approach enables automated CD activity assessment without modifying existing workflows, facilitating longitudinal monitoring, research scalability, and future integration into clinical decision support systems[135].

EMERGING AND FUTURE OPPORTUNITIES IN IBD

Advancement of precision medicine in IBD

A significant emerging opportunity in IBD research is the advancement of precision medicine through the integration of multi-omics, spatial omics, and AI, facilitating the translation of discovery-based findings into clinical practice. AI-enabled integration allows understanding of IBD as a systems-level disease, rather than treating it as a uniform condition[136]. Beyond discovering disease mechanisms and determinants of therapeutic response, AI can inform real-time clinical decision-making and patient stratification, creation of highly targeted therapies, enabling healthcare providers to choose the most effective medications and ideal dosages tailored to individual disease profiles[137].

At present, most applications of AI-driven and omics-based analyses remain confined to research settings, with limited integration into routine care and clinical trials. Broader clinical adoption will be facilitated by continued reductions in sequencing costs, streamlined sample processing, and data acquisition through advanced AI-enabled standardization and interpretability. Hence, this will accelerate the incorporation of high-dimensional molecular profiling into clinical trials and routine IBD management[138]. Additionally, AI’s ability to predict outcomes can assist in consistently identifying patients who are most likely to benefit from specific therapies, thereby decreasing unnecessary medication exposure, minimizing side effects, and lowering healthcare expenses[20].

Creation of computational modelling of tissue-specific drug responses

AI can integrate a patient’s genetic background, microbiome composition, immune profile, and spatial tissue architecture to generate a personalized modelling of tissue-specific drug responses[82,139]. The effectiveness of IBD therapies depends not only on molecular targets but also on their spatial context within inflamed tissue. AI-assisted computational modelling can assess how drug targets are distributed across distinct gut microenvironments and predict tissue-level responses to treatment[140]. A key future direction for the integration of AI with omics technologies in IBD lies in the development of next-generation, non-animal disease models that enable examination of intestinal inflammation.

Advanced in vitro systems, including intestinal epithelial cell lines, three-dimensional organoids, and microfluidic gut-on-chip platforms, are increasingly recognized as physiologically relevant and ethically sustainable alternatives for modelling IBD pathobiology[129]. These experimental platforms are further strengthened by complementary in silico approaches, such as molecular docking, network pharmacology, toxicogenomics, and AI-based modelling, which allow predictive simulations of drug target interactions and pathway-level effects. Together, these integrative frameworks facilitate the exploration of emerging disease mechanisms, including epithelial barrier modulation and multi-target inflammatory regulation, thereby accelerating translational discovery while reducing reliance on animal models[129].

Emerging opportunities in spatial host-microbiome mapping

A distinctive feature of research on the gastrointestinal tract is the close spatial relationship between host tissues and the microbiome. The spatial organization of the gut microbiota is a key determinant of microbial colonization, interspecies interactions, host-microbe crosstalk, and community stability. Dissecting these complex microbial networks requires approaches that preserve spatial context while simultaneously capturing high-resolution genetic information, enabling direct linkage between microbial localization, functional potential, and host responses[141].

Spatially resolved host microbiome profiling is emerging as a powerful approach to dissect intestinal inflammation within its native tissue. A recent multi-omics spatial framework, termed microbiome cartography, exemplifies this strategy by enabling simultaneous in situ interrogation of host and microbial features across complementary spatial modalities. Applied in a murine colitis model, this approach integrated spatial proteomics, transcriptomics, and glycomics to capture coordinated alterations in immune activation, epithelial cell states, and microbial community structure. Spatial multi-omics mapping revealed organized tissue remodeling associated with immune-epithelial perturbations, shifts in bacterial populations, inflammatory niches, and changes in metabolic pathways during disease progression[142]. Such frameworks illustrate how spatial host-microbiome technologies can provide mechanistic insight into the localized crosstalk between intestinal tissues and resident microbes, highlighting their translational potential for advancing IBD research.

Prediction of disease progression and fibrosis state

A major clinical challenge in IBD is determining when intestinal inflammation transitions into irreversible fibrosis, at which point medical therapy becomes ineffective, and surgery is unavoidable[143]. In IBD, sustained mucosal inflammation compromises epithelial barrier integrity and drives progressive injury to gastrointestinal tissues. Although fibrosis initially represents a reparative response, chronic cycles of tissue damage and wound healing in longstanding IBD promote excessive extracellular matrix deposition and pathological scarring. Intestinal fibrosis represents a major unsolved clinical challenge in IBD, substantially influencing disease trajectory and patient quality of life[144]. Spatial omics can map collagen deposition, extracellular matrix remodeling, and tissue stiffness across gut layers, while AI can analyze these spatial and molecular patterns longitudinally. This combined approach may allow clinicians to identify a predictive threshold, the fibrosis state enabling earlier intervention, improved surgical planning, and more informed patient counselling.

Emerging opportunities in the discovery of spatial biomarkers for disease stratification

Spatially resolved molecular profiling enables the identification of spatial biomarkers that reflect not only the presence of specific cell types or molecules but also their tissue localization and microenvironmental setting. A recent study engineered spatial and molecular features from defined cellular niches to inform IBD classification. By deconstructing colonic tissue into four discrete niches, the authors integrated niche composition, spatial architecture, and localized gene expression into predictive models. It was found that subtype-specific molecular programmes are embedded within spatially organized cellular niches[145].

Spatial transcriptomic analysis of mucosal biopsies from patients with postoperative CD identified conserved molecular programs associated with histologic inflammation across distinct anatomical sites, including the colon, neo-terminal ileum, and anastomosis[114]. Inflamed regions from these locations exhibited shared transcriptional signatures across multiple immune niches and a distinct epithelial cell population characterized by elevated expression of inflammation-related genes. These spatially defined inflammatory compartments were further analyzed to identify their cellular composition, gene expression profiles, and functional contributions to disease pathogenesis. Inflamed tissues across all sites showed consistent enrichment for immune cell populations and inflammatory signaling pathways, rather than hypoxia-associated programmes. In addition to canonical TNF-α and JAK-STAT signaling pathways targeted by current therapies, this study highlighted the involvement of EGFR and MAPK signaling, suggesting additional therapeutic vulnerabilities in postoperative disease recurrence[114].

In the search for spatially resolved biomarkers in IBD, spatial omics approaches have identified a distinct population of inflamed epithelial cells in CD that mirrors the transcriptional and spatial characteristics of previously described disease-associated epithelial states. This population, termed LND cells, is defined by coordinated high expression of LCN2, NOS2, and DUOX2 and is selectively enriched within inflamed colonic and ileal regions in CD[146]. Spatial validation using hybridization chain reaction FISH demonstrated co-localization of these transcripts within discrete epithelial subsets in tissues with high LND abundance. At the same time, expression was largely absent in non-IBD controls and minimal in tissues with low LND representation. Complementary protein-level confirmation by immunohistochemistry and multiplex immunofluorescence further substantiated the spatial co-expression of LCN2, NOS2, and DUOX2 within inflamed epithelial compartments[146]. Collectively, these findings establish LND cells as a spatially defined epithelial state associated with CD inflammation and highlight their potential utility as biomarkers of epithelial stress and disease activity.

Emerging application of spatial epigenomics in IBD

The adoption of spatial epigenomic approaches is expected to significantly advance IBD research by providing deeper resolution of disease-associated cellular and regulatory states while preserving native tissue architecture. Spatial epigenomics examines the spatial organization of epigenetic states within intact tissues, including chromatin accessibility, DNA methylation, histone modifications, and three-dimensional genome architecture, providing insight into how gene regulatory programs are locally established and maintained in health and disease[147].

Rapid advances in spatial omics are transforming the ability to profile cellular states within their native tissue context. Emerging methods now enable spatially resolved, genome-wide, joint profiling of the epigenome and transcriptome, allowing simultaneous sequencing of chromatin accessibility and gene expression, or histone modifications such as H3K27me3, H3K27ac, and H3K4me3, together with transcriptional output on the same tissue section at near single-cell resolution[148]. Integrative assay for transposase-accessible chromatin with high-throughput sequencing and single-cell RNA-seq profiling of intestinal and peripheral CD4⁺ T cells from patients with IBD showed that inflammation-specific chromatin accessibility programs were enriched for inflammatory gene networks. Inflamed intestinal biopsies displayed epigenetic signatures associated with preferential expansion of pathogenic Th17 cells, linking chromatin remodeling to local immune dysregulation in IBD[149].

Accelerated and inclusive data sharing for IBD research

The integration of AI with multi-omics and spatial omics technologies offers a robust framework for harmonizing and analyzing large, heterogeneous datasets across institutions and populations. AI-driven platforms can standardize data formats, reduce batch effects, and enable federated learning approaches that preserve patient privacy while promoting collaboration. Open sharing of large-scale, well-phenotyped datasets is essential for accelerating discovery in IBD research. The 1000 IBD project was established as a prospective longitudinal cohort to thoroughly profile more than 1000 individuals with IBD from the northern Netherlands. This initiative integrates extensive clinical phenotyping with multi-omics data layers, including genomic variation, intestinal transcriptomics, and gut microbiome composition[150]. Collected metadata encompass dietary environmental exposures, treatment responses, and adverse drug events, enabling systems-level analyses of disease heterogeneity. Importantly, all generated datasets are being made publicly available through established repositories, providing useful resources to support integrative and reproducible IBD research[150].

TRANSLATIONAL CHALLENGES

The path from molecular discovery to precision-guided clinical application in IBD is constrained by sequential and interdependent bottlenecks operating across biological, technological, computational, and implementation domains (Figure 2). The following section outlines the main challenges.

Figure 2 Schematic representation of the challenges that limit the translation of omics discoveries into clinically actionable precision medicine in inflammatory bowel disease. Biological heterogeneity forms the foundational layer of complexity. Multimodal data acquisition introduces additional challenges, including sampling bias, incomplete modality overlap, cross-platform variability, and multi-scale resolution mismatches. Integration of high-dimensional datasets creates a computational bottleneck driven by dimensionality overload and signal-to-noise imbalance. Artificial intelligence models face challenges related to overfitting, limited validation, and interpretability. Together, these challenges delay the realization of precision medicine outputs such as patient stratification, therapy selection, fibrosis prediction, and cancer risk mapping. A real-world data feedback loop, incorporating prospective validation and iterative recalibration, is essential to overcome some translational constraints.

Biological and spatial heterogeneity of IBD

The primary translational limitations lie in the intrinsic biological diversity of IBD. The disease is characterized by dynamic host microbiome interactions, fluctuating inflammatory activity, and pronounced microenvironmental diversity within affected intestinal segments[151]. Inflammation and fibrosis coexist along a continuum rather than representing discrete states, complicating molecular interpretation. Spatial variation within individual patients, differences in immune-cell infiltration patterns, and therapy-induced tissue remodeling contribute to substantial intra- and inter-patient variability[152]. Consequently, any single molecular snapshot captures only a context-dependent fraction of disease biology, limiting reproducibility and stability of integrative analyses. Therefore, this biological complexity represents the first major bottleneck: Molecular heterogeneity precedes and amplifies downstream computational challenges.

Multimodal data acquisition, standardization, and cross-platform variability

Beyond biological complexity, methodological constraints arise during multimodal data generation. Sampling bias related to biopsy location, tissue depth, and endoscopic accessibility can distort spatial and cellular representations of inflammation. Incomplete modality overlaps, in which not all patients undergo uniform genomic, transcriptomic, proteomic, metabolomic, and spatial profiling, reduce statistical power and complicate harmonization[153,154]. Resolution mismatches further complicate integration. Bulk tissue profiling, single-cell sequencing, imaging-based spatial data, and systemic biomarkers operate across distinct biological scales. Cross-platform differences in sequencing chemistries, imaging platforms, library preparation protocols, and preprocessing pipelines introduce batch effects and technical noise[155]. The result is a high-dimensional, partially overlapping, structurally heterogeneous dataset landscape that challenges reproducible cross-study comparisons.

Computational integration and high-dimensionality constraints

When multimodal layers are combined, a central computational bottleneck emerges. The dimensionality of integrated datasets frequently exceeds available patient numbers, creating an unfavorable feature-to-sample ratio and increasing susceptibility to overfitting[156]. Signal-to-noise imbalance, feature redundancy, collinearity across modalities, and residual batch effects can obscure biologically meaningful signals[157]. Without rigorous harmonization, principled dimensionality reduction, and robust cross-validation, integrative frameworks risk prioritizing statistical correlations over mechanistic inference. This integration bottleneck represents a critical barrier between molecular discovery and clinically actionable insight.

Algorithmic limitations and explainability in AI modelling

AI-driven modelling in IBD faces doubt due to its lack of interpretability. Although AI models can detect nonlinear relationships within complex datasets, their performance is often restricted by small or imbalanced cohorts, particularly when CD and UC phenotypes are unevenly represented[158]. Overfitting remains a significant concern, especially in the absence of independent external validation. Moreover, many AI systems function as “black boxes”, limiting interpretability and clinician trust[159]. Restricted transparency complicates regulatory approval and impedes integration into clinical workflows[160]. Even models demonstrating strong internal performance may fail in real-world settings due to distributional shifts, evolving therapeutic strategies, or unrecognized biological subtypes.

Clinical implementation, regulatory, and real-world validation barriers

Beyond computational performance, clinical implementation presents additional challenges. Comprehensive multi-omics and spatial assays remain costly and resource-intensive, limiting scalability. Turnaround times for complex analyses may not align with urgent clinical decision-making. Regulatory frameworks for AI-driven clinical decision-support tools remain evolving and heterogeneous across jurisdictions. Sustainable progress requires a learning healthcare ecosystem in which predictive models undergo prospective validation, longitudinal outcome tracking, and iterative recalibration[36]. Integration into electronic health record systems is frequently incomplete, and data privacy regulations restrict cross-border data sharing and multi-centre harmonization[161].

CONCLUSION

The integration of spatial omics, multi-omics profiling, and AI is transforming the understanding of IBD from a homogeneous inflammatory disorder to a spatially organized, systems-level condition. Multi-omics allow detailed examination of host-microbiome interactions, immune-epithelial communication, and tissue-specific inflammatory pathways that contribute to variations in disease presentation, treatment response, and long-term outcomes. To achieve reproducibility and comparability in IBD research, we propose standardized tissue sampling approaches, including the location, depth, and inflammatory status of biopsies, along with ensuring consistent processing of biospecimens, data collection, and analytical procedures. Additionally, the systematic gathering of high-quality clinical metadata, including treatment history, disease duration, endoscopic and histological findings, and pertinent environmental influences, is vital for conducting meaningful integrative analyses. We propose the use of AI in IBD must evolve beyond proof-of-concept demonstrations toward rigorous external validation across independent cohorts, platforms, and clinical settings. Emphasis on model interpretability and biological transparency will be critical to foster clinical trust, facilitate regulatory approval, and enable real-world implementation. In the near future, the most readily translatable advances are likely to emerge from AI-assisted endoscopic and histopathologic assessment, complemented by non-invasive blood or stool-based biomarkers derived from multi-omics integration. Together, these approaches provide scalable and minimally invasive tools for disease monitoring, therapeutic stratification, and response assessment, paving the way for the broader adoption of precision medicine in IBD. On the other hand, important limitations should be considered when applying AI to IBD multi-omics and spatial data, including bias introduced by non-representative cohorts, batch effects, and site-specific protocols that can reduce generalizability across populations and care settings. In addition, interpretability constraints and privacy concerns surrounding high-dimensional clinical, imaging, and omics datasets emphasize the need for transparent models, privacy-preserving governance, and rigorous external validation before widespread clinical deployment.

ACKNOWLEDGEMENTS

The authors would like to express their sincere gratitude to Dr. Pearl JE of the University of Leicester, United Kingdom, for his support in reviewing the manuscript and assisting in editing and improving the quality of the English language. Gradisteanu Pircalabioru G acknowledges the European Union’s Horizon Europe Research and Innovation framework programme 2021-2027, under the Coordination and Support Actions, funded by the European Union. The views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.