Research Radar — 2026-07-26
Methods & AI
Computational
Full-length single-cell spatial transcriptomics reveals spatial and cell-type-specific transcript isoforms in the primate brain
Nature Methods Published 2026-07-24 Research Article DOI: 10.1038/s41592-026-03174-y
spatial transcriptomics single-cell isoform resolution primate brain long-read sequencing
Summary: Presents a method combining long-read sequencing with spatial transcriptomics to profile full-length RNA isoforms at single-cell resolution in the primate brain. Reveals that spatial context and cell type jointly shape isoform usage, with hundreds of previously unannotated isoforms showing cell-type-specific and region-specific expression patterns. Demonstrates that isoform-level information captures regulatory complexity missed by gene-level spatial analyses.
Why it matters: Most spatial transcriptomics studies quantify gene-level expression, missing the functional diversity encoded by alternative splicing. Full-length isoform resolution in spatial context reveals a hidden layer of post-transcriptional regulation that could explain cell-type-specific functions and disease-associated splicing in the brain and beyond.
Why for Yiru: Directly relevant to your expertise in spatial transcriptomics and single-cell analysis. The isoform-level resolution adds a new dimension to spatial biology that could be applied to tumor microenvironment studies, where alternative splicing shapes immune recognition and cancer cell states.
RETROFIT: Reference-free deconvolution of cell-type mixtures in spatial transcriptomics
Nature Communications Published 2026-07-24 Research Article DOI: 10.1038/s41467-026-74928-7
spatial transcriptomics deconvolution cell-type mixtures reference-free computational method
Summary: Introduces RETROFIT, a computational method for deconvolving cell-type mixtures in spatial transcriptomics data without requiring matched single-cell RNA-seq references. Uses a novel statistical framework that leverages spatial autocorrelation patterns and gene expression coherence across neighboring spots to infer cell-type proportions. Demonstrates superior accuracy across diverse spatial platforms (10x Visium, MERFISH, Slide-seq) and tissue types compared to reference-based methods.
Why it matters: Most spatial transcriptomics deconvolution methods require high-quality matched single-cell references, which are often unavailable for clinical or archival samples. A reference-free approach broadens the applicability of spatial analysis to routine FFPE samples and enables retrospective analysis of archival tissue cohorts.
Why for Yiru: Reference-free deconvolution is directly useful for analyzing spatial transcriptomics data from clinical specimens where matched scRNA-seq references are unavailable. The method's compatibility with FFPE samples is particularly valuable for translational tumor microenvironment research.
Spatialproteomics: an interoperable toolbox for analyzing highly multiplexed fluorescence image data
Nature Methods Published 2026-07-24 Research Article DOI: 10.1038/s41592-026-03155-1
spatial proteomics multiplexed imaging toolbox image analysis interoperability
Summary: Presents Spatialproteomics, an interoperable computational toolbox designed for analyzing highly multiplexed fluorescence image data from platforms such as CODEX, MIBI, CyCIF, and Imaging Mass Cytometry. Provides standardized workflows for preprocessing, cell segmentation, feature extraction, spatial neighborhood analysis, and multi-sample integration. Built on a modular architecture that interfaces with popular spatial analysis frameworks including Squidpy and Giotto.
Why it matters: Highly multiplexed imaging is generating increasingly complex spatial proteomics data, but analyzing this data requires specialized computational tools that are often platform-specific. An interoperable toolbox that standardizes analysis across platforms lowers the barrier to spatial proteomics and promotes reproducible research.
Why for Yiru: Multiplexed imaging is a cornerstone technology for spatial tumor microenvironment profiling. Having access to interoperable analysis tools that integrate with your existing spatial analysis pipelines would streamline the analysis of highly multiplexed imaging data.
A benchmark study of vision and pathology foundation models for computational pathology
Nature Communications Published 2026-07-24 Research Article DOI: 10.1038/s41467-026-76004-6
foundation models computational pathology benchmark vision models deep learning
Summary: Provides a systematic benchmark comparing general-purpose vision foundation models and pathology-specific foundation models across a wide range of computational pathology tasks including tissue classification, segmentation, biomarker prediction, and survival analysis. Evaluates 15+ foundation models on 20+ public and private datasets. Reveals that while pathology-specific models show advantages on tasks requiring domain knowledge, general vision models can match or exceed them on certain tasks when properly fine-tuned, with practical recommendations for model selection based on task type and data scale.
Why it matters: Foundation models are transforming computational pathology, but the proliferation of model choices creates confusion about which to use for specific tasks. This benchmark provides evidence-based guidance for the pathology AI community, potentially saving significant compute resources and improving task-specific performance.
Why for Yiru: Foundation model selection is increasingly important for your work in spatial biology and computational pathology. Understanding when to use domain-specific vs. general vision models informs practical decisions in analyzing histological and spatial omics data.
SPgen: Proteome-wide Spatial Proteomics generation using multi-modality foundation models
bioRxiv Published 2026-07-20 Preprint DOI: 10.1101/2026.07.16.739037
spatial proteomics foundation models multi-modality protein generation deep learning
Summary: Introduces SPgen, a multi-modality foundation model framework for proteome-wide generation of spatial proteomics data. The model integrates transcriptomic, imaging, and prior protein interaction data to predict spatial protein abundance and localization across the proteome. Demonstrates the ability to generate spatial proteomic profiles from transcriptomic inputs alone, enabling spatial proteomics analysis on existing spatial transcriptomics datasets without additional wet-lab experiments.
Why it matters: Spatial proteomics provides a more direct readout of cellular function than transcriptomics, but it remains technically challenging and costly. An AI model that can generate spatial proteomic maps from transcriptomic data could dramatically expand access to proteome-level spatial information and enable re-analysis of existing spatial transcriptomics datasets.
Why for Yiru: Spatial proteomics generation from transcriptomic data is a powerful concept that could extend your spatial analysis capabilities. The multi-modality foundation model approach is directly relevant to your work in deep learning for spatial biology.
IOBRpy enables agentic multi-omics decoding of anti-tumor immunity
bioRxiv Published 2026-07-21 Preprint DOI: 10.1101/2026.07.17.739055
multi-omics anti-tumor immunity agentic AI tumor microenvironment bioinformatics
Summary: Presents IOBRpy, an agentic AI framework that integrates and interprets multi-omics data — including transcriptomics, genomics, proteomics, and spatial data — to decode anti-tumor immune responses. The platform uses LLM-powered agents to orchestrate complex analytical workflows, automatically selecting appropriate statistical methods, interpreting results, and generating hypotheses. Applied to multiple cancer immunotherapy cohorts, IOBRpy identifies known and novel biomarkers of response and resistance.
Why it matters: Multi-omics data integration for cancer immunology remains a major analytical challenge that requires expertise across multiple domains. An agentic AI system that can autonomously design and execute analytical workflows could democratize multi-omics analysis and accelerate discovery of immunotherapy biomarkers.
Why for Yiru: The agentic AI approach to multi-omics integration is highly relevant to your work in TME analysis. The platform's focus on anti-tumor immunity directly supports your research on immunotherapy response prediction and biomarker discovery.
Biomedical discoveries
Biomedicine
Disease-associated microglia adopt stage-specific phenotypes that regulate T cell fate and immunity in glioma
Immunity Published 2026-07-23 Research Article DOI: 10.1016/j.immuni.2026.06.024
microglia glioma T cell tumor microenvironment immunity brain tumor
Summary: Characterizes how disease-associated microglia in glioma transition through distinct phenotypic stages during tumor progression, each with specific functional roles in regulating T cell responses. Early-stage microglia promote T cell recruitment and activation, while late-stage microglia adopt immunosuppressive phenotypes that promote T cell exhaustion and exclusion. Identifies molecular drivers of this phenotypic switch and demonstrates that targeting the microglial transition point restores anti-tumor immunity in preclinical models.
Why it matters: Microglia are the most abundant immune cells in the brain tumor microenvironment, yet their role in regulating T cell immunity has been poorly understood. This study reveals that microglia are not passive bystanders but active, stage-dependent regulators of T cell fate — a finding with direct implications for immunotherapeutic strategies in glioblastoma.
Why for Yiru: Microglia-T cell crosstalk in the glioma TME is highly relevant to your interests in tumor-infiltrating immune cells and immunotherapy resistance mechanisms. The phenotypic plasticity of microglia parallels macrophage polarization in other cancer types, linking to your work on tumor-associated macrophages.
Localized PD-1 CAR T therapy reprograms neuroinflammation
Cell Published 2026-07-20 Research Article DOI: 10.1016/j.cell.2026.06.036
CAR-T PD-1 neuroinflammation cell therapy immunotherapy brain
Summary: Develops a localized PD-1-targeted CAR T cell therapy that is administered directly into the central nervous system to reprogram pathological neuroinflammation. Unlike systemic CAR-T approaches, localized delivery achieves high intratumoral CAR-T persistence while minimizing systemic toxicity. The PD-1-targeted CAR T cells remodel the neuroinflammatory milieu by eliminating PD-1-expressing pathogenic immune cells and shifting microglial phenotypes from pro-inflammatory to regulatory states. Demonstrates efficacy in models of neuroinflammation-associated pathology.
Why it matters: Systemic CAR-T therapy carries risks of neurotoxicity and limited CNS penetration. A localized CAR-T approach that targets PD-1-expressing cells within the brain represents a paradigm for treating CNS diseases with cell therapy, minimizing off-tumor toxicity while achieving therapeutic efficacy in a compartmentalized manner.
Why for Yiru: Directly relevant to your interest in CAR-T cell therapy and tumor microenvironment engineering. The concept of localized cell therapy delivery and PD-1 targeting could be translated to brain tumors, where CAR-T approaches are actively being developed.
Single-nucleus multimodal spatial transcriptomics reveals spatial colocalization of neoantigen-expressing tumor cells and cognate T cells
Nature Biotechnology Published 2026-07-22 Research Article DOI: 10.1038/s41587-026-03194-1
spatial transcriptomics single-nucleus neoantigen T cell tumor microenvironment immunotherapy
Summary: Develops a multimodal spatial transcriptomics approach that simultaneously profiles transcriptomes at single-nucleus resolution and maps T cell receptor (TCR) sequences within intact tissue sections. Applied to human tumors, the method enables the first direct visualization of neoantigen-expressing tumor cells spatially colocalized with their cognate T cells. Reveals that productive anti-tumor immunity requires spatial proximity between neoantigen-presenting tumor cells and TCR-matched T cells within specific tissue microenvironments, and that this spatial organization is disrupted in non-responders to immunotherapy.
Why it matters: Understanding the spatial relationship between tumor neoantigens and cognate T cells is fundamental to cancer immunotherapy. This study provides the first direct evidence that spatial colocalization of neoantigen-expressing tumor cells with their cognate T cells is a determinant of effective anti-tumor immunity, with immediate implications for biomarker development and therapeutic strategy.
Why for Yiru: This is remarkably aligned with your research at the intersection of spatial transcriptomics and cancer immunotherapy. The ability to map TCR specificity alongside spatial transcriptomic profiles opens up new avenues for understanding how spatial organization of the immune microenvironment influences immunotherapy response.
Tertiary lymphoid structures harbour stem-like tumour-specific T cells
Nature Published 2026-07-22 Research Article DOI: 10.1038/s41586-026-10808-w
tertiary lymphoid structures T cell stem-like T cells tumor immunity immunotherapy
Summary: Provides a comprehensive characterization of T cells within tertiary lymphoid structures (TLS) in human tumors. Using single-cell RNA-seq, TCR-seq, and spatial analysis, demonstrates that TLS are enriched for stem-like TCF1+PD-1+ tumor-specific T cells that retain proliferative capacity and multipotency. These TLS-resident stem-like T cells are distinct from exhausted T cells in the tumor core and can give rise to effector T cells that mediate anti-tumor responses. The presence and maturity of TLS correlates with improved immunotherapy outcomes across multiple cancer types.
Why it matters: TLS are increasingly recognized as predictive biomarkers for immunotherapy response, but their functional role has been unclear. This study establishes TLS as specialized niches that maintain a reservoir of stem-like tumor-specific T cells, explaining why TLS-positive tumors have more durable responses to checkpoint blockade and providing a rationale for therapeutic TLS induction.
Why for Yiru: TLS biology is a rapidly emerging area in cancer immunology with direct implications for biomarker development and therapeutic targeting. The stem-like T cell niche concept connects to your interests in T cell dysfunction and exhaustion within the TME.
Macrophage-instructed GSDME couples glioblastoma cell-state plasticity with inflammatory cell death
bioRxiv Published 2026-07-22 Preprint DOI: 10.1101/2026.07.22.740157
macrophage GSDME glioblastoma cell-state plasticity inflammatory cell death tumor microenvironment
Summary: Reveals that tumor-associated macrophages instruct glioblastoma cells to express GSDME through paracrine signaling, coupling cell-state plasticity with inflammatory cell death. Macrophage-derived signals promote a mesenchymal-like transition in glioma cells that upregulates GSDME, rendering them susceptible to pyroptosis upon cytotoxic stress. This macrophage-glioma axis creates a feedback loop where inflammatory cell death further propagates macrophage activation, shaping the overall tumor immune landscape.
Why it matters: Pyroptosis and inflammatory cell death are emerging as key determinants of anti-tumor immunity, but the mechanisms regulating GSDME expression in cancer cells have been unclear. The discovery that macrophages actively instruct glioma cell-state transitions to control pyroptotic susceptibility reveals a new dimension of immune-epithelial crosstalk in the TME.
Why for Yiru: The macrophage-glioma crosstalk mechanism is directly relevant to your research on tumor-associated macrophages and their role in shaping the TME. The connection between cell-state plasticity and inflammatory cell death provides a new framework for understanding how macrophage polarization influences tumor cell fate.
Cross-disciplinary watchlist
Other Fields
Efficient and precise programmable DNA knock-in without double-strand breaks
Nature Published 2026-07-22 Research Article DOI: 10.1038/s41586-026-10819-7
gene editing DNA knock-in double-strand break-free CRISPR genome engineering
Summary: Develops a programmable method for precise DNA knock-in that bypasses the need for double-strand breaks (DSBs), a major source of genotoxicity in current genome editing approaches. The method uses a CRISPR-guided nickase combined with a novel DNA repair template design that leverages alternative repair pathways to achieve efficient and precise insertion. Demonstrates knock-in efficiency comparable to or exceeding DSB-dependent methods across multiple cell types and loci, with significantly reduced rates of large deletions, translocations, and p53 activation.
Why it matters: Double-strand breaks are the primary source of genotoxicity in CRISPR-based genome editing, limiting therapeutic applications. A knock-in method that avoids DSBs could dramatically improve the safety profile of gene therapies — particularly important for in vivo editing where off-target effects are harder to control — and expand the scope of precisely editable cell types.
Why for Yiru: Gene editing is a foundational technology for functional genomics and potential therapeutic applications. DSB-free knock-in methods could enable safer engineering of immune cells for adoptive cell therapy and facilitate more precise functional studies of genes in the TME.
Spatial multi-omics identifies early synaptic pruning and context-specific dopaminergic vulnerability in synucleinopathies
Nature Communications Published 2026-07-21 Research Article DOI: 10.1038/s41467-026-74961-6
spatial multi-omics synucleinopathy Parkinson's disease synaptic pruning dopaminergic neurons
Summary: Uses spatial multi-omics — combining spatial transcriptomics, proteomics, and metabolomics — to map the molecular landscape of synucleinopathies at unprecedented resolution. Identifies early synaptic pruning as an initiating event that precedes overt neuronal loss, and reveals that dopaminergic neuron vulnerability is highly context-dependent, shaped by local glial interactions and metabolic environment. Provides a spatial atlas of disease progression that identifies early intervention windows.
Why it matters: Synucleinopathies like Parkinson's disease are diagnosed late, after significant neuronal loss has occurred. Spatial multi-omics reveals the earliest molecular events — including synaptic pruning — that precede neuronal death, identifying potential biomarkers for early diagnosis and therapeutic targets for disease modification before irreversible damage occurs.
Why for Yiru: Spatial multi-omics approaches that integrate transcriptomics, proteomics, and metabolomics are directly relevant to your spatial biology expertise. The methodological framework for integrating multi-modal spatial data is transferable to tumor microenvironment studies.
Spatial transcriptomics reveals distinct cell type dynamics following opioid dependence in female mice with the common human μ-opioid receptor variant Oprm1 A118G
Nature Communications Published 2026-07-23 Research Article DOI: 10.1038/s41467-026-75723-0
spatial transcriptomics opioid dependence μ-opioid receptor cell type dynamics neuroscience
Summary: Applies spatial transcriptomics to map cell-type-specific transcriptional responses to opioid dependence in female mice carrying the common human OPRM1 A118G variant. Reveals distinct spatial patterns of neuronal and glial activation across brain regions, demonstrating that the A118G variant significantly alters the cellular response to chronic opioid exposure. Identifies region-specific microglial and astrocytic activation states that correlate with behavioral phenotypes of dependence.
Why it matters: The OPRM1 A118G variant is carried by approximately 30% of the human population and influences opioid response and addiction risk, yet its effects on brain cell types at spatial resolution have been unknown. This study provides the first spatially resolved view of how a common genetic variant shapes the cellular response to opioids.
Why for Yiru: Spatial transcriptomics in neuroscience applications demonstrates the versatility of spatial techniques beyond cancer biology. The methodological approach to mapping cell-type-specific responses in a genetically modified model is transferable to studies of the TME and immunotherapy response.
Subnuclear genome compartmentalization controls bivalent chromatin activity
Nature Published 2026-07-22 Research Article DOI: 10.1038/s41586-026-10832-w
chromatin 3D genome bivalent chromatin gene regulation nuclear organization
Summary: Investigates how subnuclear genome compartmentalization regulates the activity of bivalent chromatin domains — regions marked by both activating H3K4me3 and repressive H3K27me3 modifications that poise developmental genes for activation. Using super-resolution imaging, Hi-C, and epigenomic profiling, demonstrates that bivalent domains are spatially compartmentalized within the nucleus, and that their repositioning between repressive and active compartments controls gene activation during differentiation. Disruption of this spatial organization leads to premature or ectopic gene activation.
Why it matters: Bivalent chromatin domains are essential for maintaining developmental genes in a poised state, and their dysregulation is implicated in cancer and developmental disorders. This study reveals that subnuclear spatial positioning is an additional layer of regulation — beyond histone modifications — that controls whether bivalent genes are activated or silenced.
Why for Yiru: 3D genome organization and chromatin regulation are fundamental to understanding how gene expression programs are controlled in development and disease. The spatial compartmentalization concept parallels the spatial organization principles you study at the tissue level.
Uni-TINT: Unifying T and Innate Lymphoid Cell Taxonomy with a high resolution, pan-disease and pan-tissue single-cell transcriptomic atlas
bioRxiv Published 2026-07-22 Preprint DOI: 10.1101/2026.07.20.739448
single-cell T cell innate lymphoid cell atlas taxonomy pan-disease
Summary: Presents Uni-TINT, a high-resolution single-cell transcriptomic atlas that unifies the taxonomy of T cells and innate lymphoid cells (ILCs) across multiple tissues, diseases, and species. Integrates over 1 million cells from diverse datasets to establish a harmonized cell-type classification that resolves long-standing ambiguities in ILC and T cell subset definitions. Provides a reference framework for annotating lymphoid cells in single-cell studies and reveals conserved and tissue-specific gene expression programs.
Why it matters: The classification of T cells and ILCs has been inconsistent across studies, tissue types, and diseases, hindering cross-study comparisons and meta-analyses. A unified, high-resolution taxonomy that spans tissues and disease contexts provides the field with an essential reference for consistent cell annotation and facilitates integration of findings across immunological studies.
Why for Yiru: A harmonized T cell and ILC taxonomy is directly useful for your single-cell RNA-seq analyses of the tumor microenvironment. Accurate cell-type annotation is foundational for downstream analyses including spatial mapping, cell-cell interaction inference, and functional characterization of tumor-infiltrating lymphocytes.