Biology subjects

Sarkar, D.

Publications and source records attributed to Sarkar, D..

At least 19 recordsLinked to original sources

bioRxiv2026

RadMultiBench: A Reproducible Dual-Task Benchmark Exposing the Limits of Pre-Trained Vision-Language Models Across Thoracic and Breast Imaging

We present RadMultiBench, a multimodal benchmark of 6,148 medical images paired with expert radiological impressions across five categories--Breast Cancer, Lung Benign, Lung Malignant, Pneumonia, and Tuberculosis--targeting two core clinical tasks: automated report generation and disease classification. To enable reliable, repeatable experimentation, we provide a reproducible pipeline for data ingestion, de-identification, split management, and standardized reporting of text and classification metrics, facilitating dataset versioning and continuous benchmark testing. We compare three architectures: (1) a classical CNN-RNN with Bahdanau attention (ResNet-LSTM), (2) a general-domain vision-language (V-L) prefix-projection model (CLIP-GPT), and (3) a domain-specific V-L prefix-projection model (BioViL-T). Results reveal a sharp performance divide: the ResNet-LSTM attains near-perfect classification (99.5% accuracy) and is the only model to generate coherent reports (BLEU-4 = 0.021), whereas both prefix-projection models catastrophically fail generation (BLEU-4 [≤] 0.0001); moreover, the domain-specific BioViL-T also fails classification (11.5% accuracy). From a data-engineering angle, the pipeline is productionizable with lineage, feature registries, and continuous integration to guard against silent regressions when updating models or data; from a data-science angle, the benchmark provides baselines and analyses that foreground feature alignment and architectural suitability as first-order concerns, inviting ablations on multimodal objectives, retrieval-augmented encoders, and uncertainty-aware decoding. Overall, RadMultiBench highlights that state-of-the-art status in general V-L domains does not translate automatically to high-stakes clinical tasks, and it furnishes the tooling and evidence base needed to develop robust, clinically viable multimodal systems.

bioinformatics↗

bioRxiv2026

Inflammation is an essential trigger for POLE-driven ultramutated colorectal cancer

Pathogenic variants in the exonuclease (proofreading) domain of DNA polymerase {varepsilon} (POLE) are frequently detected in colorectal cancers (CRC) with high tumor mutational burden (TMB), yet their functional role in intestinal tumorigenesis remains unknown. Here, we investigated the impact of mutation in POLE on intestinal homeostasis using a conditional mouse model expressing the PoleP286R pathogenic driver in intestinal epithelial cells. PoleP286R mutant mice developed invasive adenocarcinomas in the small intestine with incomplete penetrance and extremely long latency but did not spontaneously develop colorectal tumors. However, induction of colitis with dextran sodium sulfate (DSS) resulted in a striking phenotypic shift to colorectal adenocarcinomas at high frequency. Molecular and histological analyses revealed hyperactivation of Wnt/{beta}-catenin, NF-{kappa}B, ERK and STAT3 pathways, accompanied by markedly higher proliferation of intestinal epithelial cells, in DSS-treated PoleP286R mutant mouse colons. Whole genome sequencing of the tumor demonstrated exceptionally high TMB (>100 mutations/megabase) and characteristic trinucleotide signatures, consistent with human POLE tumors. Recurrent mutations in Apc and Ctnnb1 underscored the central role of Wnt signaling in POLE-driven hypermutated CRC. In addition, intestinal organoids derived from PoleP286R mutant mice exhibited accelerated growth and increased expression of pro-proliferative genes. Together, these findings highlight POLE as a gatekeeper of intestinal genomic integrity and identify inflammation as the key driver that unleashes POLE-mutant CRC. Furthermore, this work establishes a mechanistically faithful preclinical platform for studying the molecular pathogenesis of CRC and advancing therapeutic drug discovery.

cancer biology↗

bioRxiv2026

Reduced phosphatases and hyperphosphorylated glutamatergic receptors persist from early to late stages of prion disease

Synapse loss occurs early in prion disease, yet the intracellular signaling driving this dysregulation remains poorly understood. Evidence of neuronal hyperactivity was reported in humans with sporadic and familial CJD as well as in the early stages of prion disease in mice (strain 22L). Here we perform a longitudinal study of prion-infected mice (strain RML) and show an increase in the immediate-early gene, Arc, suggesting this is a conserved response to prions. We also observed a persistent decrease in phosphatase abundance (PP1 and PP2B) starting at the earliest timepoint (preclinical) in the cerebral cortex. At this stage, mGluR5 dimers and phosphorylated AMPA receptors were increased. By clinical disease onset, mGluR5 dimers declined, and activity shifted toward NMDA receptor activation, marked by GluN2B hyperphosphorylation. Together, these findings suggest that early phosphatase reduction accompanies neuronal hyperactivity and precedes substantial inflammation. Elucidating the mechanisms driving this early phosphatase loss and its role in neuronal hyperactivity may reveal novel therapeutic targets for prion disease.

neuroscience↗

bioRxiv2026

A Lenalidomide-Inducible Suicide Switch for Gene- and Cell-Based Therapy

Genetically modified immune effector cells, including chimeric antigen receptor (CAR) T cells, have emerged as transformative anti-cancer agents. However, these autonomous, living therapies are limited by difficult-to-control hyperactivation toxicities. To achieve their full potential as therapies for a broader range of diseases, CAR T cells must target novel antigens and integrate further potency enhancements that raise the possibility of novel and intensified toxicities. High performance safety systems are needed to restrain these highly active next-generation gene- and cell-based therapies. Cell therapy suicide switches remain infrequently used, and are currently limited by inaccessible controller drugs, incomplete cell depletion, fitness cost, and immunogenicity. To address these issues, we engineered a fully human sequence-based cell therapy suicide switch based on the overexpression of Caspase-Activated DNase (CAD) and its stoichiometric chaperone/inhibitor ICAD fused to a lenalidomide-inducible degron. Acute destabilization of ICAD with the broadly available, off-patent molecular glue degrader drug lenalidomide uncaged CAD to induce DNA damage and cell death. By this degradable cage mechanism, the ICAD-CAD suicide switch achieved long-term expression in primary human T cells without detectable leakiness, and deep drug-induced cell depletion. The ICAD-CAD switch was generalizable to alternative acute destabilization systems. Genome-wide CRISPR screening established the necessity of CAD and the lenalidomide-interacting E3 ubiquitin ligase substrate adaptor CRBN for switch function. When co-delivered with a CAR, ICAD-CAD CAR T cells retained effector functions and could be rapidly depleted with lenalidomide.

synthetic biology↗

bioRxiv2026

Lipid-Protein Reciprocal Coupling: A Quantitative Analysis of Membrane Curvature, Architecture, and Protein Association

Cellular membranes are dynamic molecular landscapes in which lipid composition, transbilayer asymmetry, membrane curvature, thickness and mechanics collectively shape protein organization and function, while proteins can in turn remodel lipid organization and membrane architecture. Yet lipid-protein interactions are often treated as discrete binding events, obscuring the extent to which the physical organization of the membrane itself constitutes an active determinant of protein conformational states and cellular activity. Here, we investigate lipid-protein reciprocal coupling by integrating quantitative analyses of curvature-dependent protein-membrane association and hydrophobic-run organization with established structural, lipidomic and biochemical evidence. Our analysis reveals substantial variation in protein association with curved versus flat membrane environments and highlights how membrane curvature, together with lipid composition, cholesterol and asymmetry, can regulate protein structure and function, including in disease-associated systems. We further examine emerging lipid sensors, spatial proteomics and advanced biomimetic membrane platforms that are beginning to bridge molecular mechanisms with protein-lipid organization in cells. In particular, suspended membranes, pore-spanning systems and nanopillar architectures offer opportunities to independently control lipid composition, leaflet asymmetry, curvature and mechanics while directly interrogating protein behaviour. Together, these findings support a model in which membrane architecture encodes physical information that is read by proteins and reciprocally reshaped by protein activity to regulate cellular function.

biophysics↗

bioRxiv2026

Discovery of a Small-Molecule mTORC1 Pathway Activator that Enhances Adoptive T-Cell Therapy via Immuno-metabolic Reprogramming of CD8+ T Cells

Adoptive cell therapy (ACT) has transformed cancer immunotherapy; however, its clinical efficacy remains limited by progressive T cell dysfunction and inadequate metabolic fitness acquired during ex vivo expansion and within the tumor microenvironment. Here, we report the discovery of AS-3, a first-in-class small-molecule activator that enhances the therapeutic competence of CD8 T cells through RAPTOR-dependent mTORC1 activation. Identified by phenotypic screening of an in-house library of pharmacologically relevant scaffolds and subsequent medicinal chemistry optimization, AS-3 markedly increased effector cytokine production while preserving T cell viability. Mechanistically, AS-3 enhanced phosphorylation of mTORC1 downstream effectors, promoted glycolytic and mitochondrial metabolism, and sustained pathway activity under rapamycin-mediated inhibition. Transcriptomic profiling of activated human CD8 T cells revealed coordinated enrichment of mTORC1 signalling, oxidative phosphorylation, glycolysis, proliferative programs, and cytotoxic effector gene networks, consistent with comprehensive immunometabolic reprogramming. Genetic silencing of RAPTOR attenuated both mTORC1 signalling and cytokine induction, establishing pathway dependency. Ex vivo conditioning with AS-3 significantly improved antitumor efficacy in two independent adoptive T-cell therapy models, accompanied by enhanced persistence, reduced exhaustion, and superior effector function. Together, these findings establish pharmacological activation of the RAPTOR-mTORC1 axis as a strategy to generate metabolically resilient T cells and provide a clinically translatable approach to improving the durability and efficacy of next-generation ACT.

immunology↗

bioRxiv2026

Deep-Interact Studio: An Interactive Deep Learning Model Building Platform for Biomolecular Interaction Prediction

MotivationDeep learning has rapidly become essential for predicting biomolecular interactions; however, most web-tools expose only a single, pre-built model with a fixed, non-configurable architecture that users cannot redesign, retrain on their own data, or compare; they are typically dedicated to one interaction type and often one species, and report prediction scores with little interpretability. These constraints force researchers across several disconnected, single-purpose tools and limit the flexibility, reproducibility, and long-term usability of existing platforms. ResultsWe present Deep-Interact Studio, a unified, web-based deep-learning platform that addresses these limitations by shifting interaction prediction from a model-centric to a user-driven, comparative, and interpretable paradigm. Within a single interface spanning all four interaction classes, namely protein-protein, drugtarget, RNA-protein, and protein-DNA, users design their own model architectures layer by layer, configure training hyperparameters, and train them on their own data, including custom, species-specific datasets. Multiple user-built models can then be trained under identical conditions and compared side by side at both the training and inference levels, while integrated interpretability, including SHAP-based feature attribution, embedding-space visualization, and interaction hub analysis, turns predictions into auditable, mechanistically grounded results. Deep-Interact Studio is, to our knowledge, the only such platform to combine fine-grained per-layer model customization with multi-model comparison and interpretability, offering a flexible and transparent alternative to fixed, single-purpose tools. Availability and implementationDeep-Interact Studio is freely available as a web application at https://deepinteract.compbiosysnbu.in/, with no login or installation required.

bioinformatics↗

bioRxiv2026

MoE-Bind: Guiding De Novo Protein Binder Generation with Sparse Experts

De novo protein binder design has been dominated by structure-based pipelines that require known three-dimensional target conformations and consume substantial compute and generation time per design, limiting their throughput and accessibility for routine large-scale binder exploration. Sequence-only generative models promise a faster and lighter alternative, yet existing systems remain uniformly dense and frequently reintroduce structural computation at inference, undermining the core advantages they were intended to deliver. Across the broader language modelling community, transformers have meanwhile transitioned from fully dense designs to sparse Mixture-of-Experts architectures that decouple capacity from per-token compute, a shift that has yet to reach sequence-only protein binder generation. We present MoE-Bind, an autoregressive protein binder generator that, for the first time in this domain, combines Multi-head Latent Attention with a sparse Mixture-of-Experts feed-forward network and is evaluated under two independent structure predictors, Boltz-2 and AlphaFold2-Multimer. Despite activating less than half the per-token parameters of compute-matched dense baselines, MoE-Bind matches or exceeds them on full-length receptor-conditioned binder generation on a leakage-free Docking Benchmark 5.0 evaluation, transfers without peptide-specific training to short-peptide design, and reduces training and inference compute by a large margin. Routing analysis on generated binders reveals interpretable expert specialization at both the individual amino acid and biochemical group level, a structured expert-token alignment not previously reported for natural-language MoE models. These results show that sparse architectural design, rather than scale, can deliver fast, structure-free, and interpretable protein binder generation.

bioinformatics↗

bioRxiv2026

Simple synthesis and functionalisation of α-hydroxyglycine-containing peptide fragments

We report here a method for the chemical synthesis of Fmocprotected -hydroxyglyine (-OH-Gly) dipeptides. Our method features operational simplicity and is compatible with protecting groups for peptide synthesis. Utility is then demonstrated through substitution at the -OH-Gly position to yield myriad non-natural amino acid-containing dipeptide fragments.

biochemistry↗

bioRxiv2026

ARACoFusion: Uncertainty-aware calibrated deep learning for protein-protein interaction network prediction in Arabidopsis thaliana

Accurate mapping of the Arabidopsis thaliana protein-protein interaction (PPI) network is essential for deciphering complexity of plant systems biology. Here, we present ARACoFusion, a specialized deep learning architecture designed to predict inter-protein connectivity directly from primary sequences. To capture the asymmetric dependencies between plant proteins, the framework utilizes a reciprocal cross-attention encoder combined with latent interaction projections and multi-source feature fusion. Addressing the severe class imbalance inherent in plant interactomes, the model integrates uncertainty-aware variance regularization and focal loss with label smoothing, further enhancing reliability through post-hoc probability calibration via temperature scaling. Extensive benchmarking on gold-standard Arabidopsis datasets demonstrates that ARACoFusion significantly outperforms existing plant-specific predictors, achieving superior scores in Area Under the Precision-Recall Curve (AUPRC), Balanced Accuracy, and Matthews Correlation Coefficient (MCC). Additionally, the model exhibits robust cross-species generalization and clear class separability in t-SNE latent space visualizations. To facilitate community-wide usage, we provide a dedicated web server for scalable network-level inference at https://ARAcofusion.compbiosysnbu.in/.

bioinformatics↗

bioRxiv2026

PFOS Disrupts Membrane Signaling and Epithelial Integrityin Fallopian Tube Cells

Perfluorooctane sulfonic acid (PFOS), a per- and polyfluoroalkyl substance (PFAS), is a widespread persistent environmental pollutant that has been implicated in various human health conditions, including infertility and cancer. Here, we investigate the effects of acute exposure to PFOS on human fallopian tube epithelial (FNE) cells that are essential for fertility and increasingly recognized as the origin site for high-grade serous ovarian cancer. We show that acute PFOS exposure changes morphology, arrests proliferation, impairs adhesion, and compromises epithelial integrity of FNE cells. Using transcriptomic profiling of FNE cells exposed to PFOS, we found increased expression of genes associated with stress-response signal transduction, including KRAS, and decreased expression of genes related to cholesterol transport and lipid homeostasis. We show that inhibition of MEK/ERK or cholesterol supplementation rescued changes in cell morphology. Further, we performed membrane fluidity measurements of cells exposed to PFOS and found elevated membrane disorder and fluidity. Our results are consistent with a model in which PFOS perturbs plasma membrane, activates stress-response signaling pathways, and impairs epithelial cell function. These studies establish a framework for understanding the effects of PFAS on cell physiology.

cell biology↗

bioRxiv2026

Elevated YBX1 mRNA expression is associated with a genomi-cally unstable and clinically aggressive cancer state: a pan-cancer analysis

Y-box binding protein 1 (YB-1; YBX1) is a multifunctional DNA- and RNA-binding protein involved in cell cycle regulation, DNA repair, stress adaptation, and therapy resistance. Elevated YBX1 mRNA expression is associated with aggressive disease across multiple cancers, yet its pan-cancer genomic and clinical correlates remain unclear. Here, we performed a comprehensive pan-cancer analysis across 53 datasets spanning 33 tumour types, integrating RNA expression, somatic mutations, copy number, hypoxia, and clinical outcomes. YBX1 was rarely mutated or amplified, indicating that oncogenic relevance is primarily driven by its expression. Tumours with high YBX1 mRNA exhibited a conserved transcriptional program enriched for cell cycle, DNA repair, and chromatin regulation pathways, and were preferentially mutated in genes involved in maintaining genomic stability, including TP53. These tumours were associated with increased mutation burden, fraction of genome altered, homologous recombination deficiency, and elevated hypoxia. Clinically, high YBX1 mRNA associated with advanced stage, higher grade, shorter progression-free survival, and reduced overall survival. Collectively, high YBX1 mRNA expression defines a conserved, genomically unstable, and clinically aggressive tumour state across multiple cancer types.

cancer biology↗

bioRxiv2025

Simulating C3-to-CAM transition under ancient atmospheric conditions

Crassulacean Acid Metabolism (CAM) plants have very low water demand and can survive in arid environment. It is hypothesized to evolve from C3 metabolism about 20 to 30 million years ago (MYA) when the Earth faced declining atmospheric CO2 concentration ([CO2]a), increasing aridity and decreasing temperature. Understanding whether and how these atmospheric changes favour C3-to-CAM transition will help in bioengineering CAM into crop plants to tackle the threat of global climate change and ensure food security. Our simulations of plant cellular metabolism under the changing atmospheric conditions of ancient times observe C3-to-CAM transition, capture possible temporal alterations of active metabolic pathways and further confirm that both the reduced [CO2]a and increased water scarcity associated with higher aridity, can act as evolutionary agents, driving the C3-to-CAM transition. Although the predicted elevated [CO2]a of future reveals a reversion towards C3-like behaviour, drought always favours CAM regardless of [CO2]a and temperature levels. Moreover, a minimum oxygen concentration is required to sustain elevated nocturnal respiration necessary for CAM, reflected by the increased activities of enzymes involved in the TCA cycle and the mitochondrial electron transport chain.

plant biology↗

bioRxiv2025

Repeat Systemic Delivery of Cross-Neutralization Resistant Synthetic Vesiculoviruses Immunomodulates the Tumor Microenvironment

The clinical efficacy of systemic oncolytic virotherapy (OV) is constrained by the rapid development of neutralizing antibodies (nAbs), which prevent repeat systemic administration, a critical barrier to sustained anti-tumor immunity. Vesiculoviruses offer potent oncolytic and immunogenic potential. However, leveraging their serological diversity for repeat dosing remains unexplored. We generated a library of chimeric vesiculovirus vectors incorporating glycoproteins from less well characterized vesiculovirus species. We evaluated vector replication, infectivity, interferon (IFN) responses, and oncolysis in vitro, alongside assessments of neutralization resistance using patient sera, monoclonal antibodies, and in silico structural modeling. In vivo studies assessed tumor delivery, immune activation, and therapeutic efficacy following intravenous administration. The vesiculovirus library exhibited broad tumor infectivity, distinct IFN-stimulatory profiles, and variable oncolytic activity. Neutralization assays and computational modeling identified serological distinctness across vectors, driven by hypervariable glycoprotein epitopes, enabling evasion of cross-neutralizing antibodies. Tumor delivery and anti-tumor immunity were preserved despite humoral barriers. Incorporating tumor-associated antigens (TAAs) further amplified anti-tumor responses, even in the context of anti-viral memory. Sequential administration of distinct vesiculovirus vectors induced robust immune activation and improved survival in a B16-OVA-IFNAR-/- model. Our findings establish a glycoprotein-diverse vesiculovirus platform capable of overcoming humoral immunity, enabling repeat intravenous dosing and sustained engagement of the tumor microenvironment. This strategy advances the field of oncolytic virotherapy by addressing a major translational barrier and lays the groundwork for future clinical studies integrating multi-vector, multi-dose immunovirotherapy with immune checkpoint blockade. One Sentence SummaryA glycoprotein-engineered vesiculovirus platform circumvents neutralizing antibodies, enabling repeat intravenous dosing and sustained anti-tumor immunity in preclinical models.

molecular biology↗

bioRxiv2025

Predictive Analysis of Methylation Patterns in Oral Squamous Cell Carcinoma (OSCC) Using Machine Learning

BackgroundOral and oropharyngeal cancers are the most common types of head and neck cancers, with over 90% originating from squamous cells in the mouth and throat. Chronic tobacco and alcohol use, inflammation, viral infections, betel quid chewing, and genetic predisposition are major risk factors for OSCC, which kills over 100,000 patients annually. Epigenetic mechanisms, such as DNA methylation, can silence tumour suppressor genes, contributing to cancer progression and patient outcomes in oral squamous cell carcinoma (OSCC). ObjectiveThis study aimed to predict prominent methylation signatures that can distinguish OSCC from normal cells. MethodsMachine learning algorithms like Support Vector Machine (SVM), Random Forest (RF), and Multilayer Perceptron (MLP) were implemented using R packages and a balanced training dataset consisting of M-values of methylated CpG sites from 46 matched OSCC and normal adjacent tissue samples. ResultsMLP model demonstrated the highest accuracy of 92% on the training dataset and 100% on the blind dataset, even with a reduced feature set of just 10 significantly differentially methylated CpG sites. ConclusionA highly accurate and generalizable machine learning model was developed using the Multi-Layer Perceptron with multiple layers (MLP-ml) algorithm, which achieved an accuracy of 95% on an independent validation dataset of 15 OSCC tumors and 7 non-tumor adjacent tissue samples.

bioinformatics↗

bioRxiv2025

Uncovering asparagine utilization pathway, a critical requirement for mycobacterial pH-driven adaptation at the host-pathogen crossroads

M. tuberculosis, an intracellular pathogen, thrives in a specialized membrane-bound vacuole, the phagosome with acidic pH and limited access to nutrients. To survive and replicate within human host, M. tuberculosis must adapt and fulfil its nutritional requirements. To understand how the bacillus captures nutrients from its host, we studied mycobacterial acquisition of host-derived amino acids, the preferred nitrogen sources. We discovered that PhoP, a key determinant of mycobacterial adaptation to phagosomal acidification, controls expression of AnsP1 and AnsP2 to facilitate acquisition of host aspartate and asparagine, respectively. Thus, macrophage-infected WT-H37Rv showed a significantly higher level of intra-bacterial Asn compared to the phoP-KO mutant and a complemented mutant could restore Asn level to that of WT-H37Rv. Under acidic conditions, elevated DNA binding of PhoP within the promoters lead to direct activation of ansP1 and ansP2. Consistently, phoP-KO is unable to utilize Asn under acidic condition, and over-expression of ansP1 or ansP2 restored intracellular growth defect of the mutant. These findings uncover the regulatory network allowing utilization of organic nitrogen sources by the pathogen during infection. Author summaryHuman Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a global public health burden claiming [~]1.25 million deaths in 2023. To investigate how the bacillus captures nitrogen sources from its host, we studied acquisition of host amino acids aspartate (Asp) and asparagine (Asn) by the tubercle bacilli. We show that PhoP, a key determinant of mycobacterial adaptation to phagosomal acidification, controls expression of Asp and Asn transporters to facilitate acquisition of host amino acids by the pathogen. A mutant lacking phoP is unable to utilize Asn under acidic conditions, whereas over-expression of ansP1 and/or ansP2 restored growth defect of the phoP-KO in vitro and in macrophages. Together, these findings uncover the transcriptional landscape allowing utilization of organic nitrogen sources by the tubercle bacilli during infection.

microbiology↗

bioRxiv2025

In-Silico Characterization of TP53 Splice Mutations in Somatic and Germline Tumours

TP53 undergoes alternative splicing to produce multiple mRNA transcripts and protein isoforms, yet the effects of splice site mutations on isoform regulation, tumor-biology, and clinical outcome remain unclear. Analysis of 23,017 TP53 variants, including 18,562 somatic mutations (pan-cancer datasets - cBioPortal) and 4,455 germline variants (IARC database), identified recurrent donor (X32, X125, X224, X261, X331) and acceptor (X33, X126, X187, X225, X307, X332) splice site mutations. Germline variants showed nucleotide-specific transition biases. Most splice site mutations were associated with reduced TP53 mRNA expression; however, X32, X33, X126, and X261 maintained or elevated transcript levels. Splice mutations were associated with distinct transcriptional subsets marked by altered p53 target gene expression, elevated tumor mutation burden, increased genomic instability, and significantly reduced disease-free survival compared to missense mutations, with X126 and X331 being associated with poorest outcomes. These findings emphasize the clinical impact of TP53 splice site mutations and the need for functional classification.

cancer biology↗

bioRxiv2025

KODAMA enables self-guided weakly supervised learning in spatial transcriptomics

Spatial transcriptomics provides researchers with a powerful tool to investigate gene expression patterns within their native positional context in tissues, revealing intricate cellular relationships. However, analyzing spatial transcriptomics data presents unique challenges due to their high dimensionality and complexity. Several competitive tools have emerged, each aiming to integrate spatial dependencies into their workflow. Yet, the overall performance of these approaches remains constrained by limitations in prediction accuracy and compatibility with specific data types. Here, we introduce the third version of the KODAMA algorithm, specifically tailored for spatial transcriptomics analysis. This upgraded version incorporates a novel approach that effectively reduces data dimensionality while preserving spatial information. At its core, the KODAMA algorithm employs parallel iterations that enforce spatial constraints throughout an embedded clustering process. Our method provides the flexibility to simultaneously analyze multiple samples, streamlining analysis workflows. Additionally, it extends beyond traditional 2D analysis to accommodate multidimensional datasets, including 3D spatial information. KODAMA seamlessly integrates into various pipelines, such as Seurat and Giotto. Extensive evaluations of KODAMA on 10x Visium, Visium HD, and image-based 3D datasets demonstrate its high accuracy in spatial domain prediction. Comparative analyses against alternative dimensionality reduction techniques and spatial analysis tools consistently validate and highlight KODAMAs superior performance in unraveling the spatial organization of cellular components across both single and integrated tissue samples.

bioinformatics↗