Research Radar — 2026-05-25
Methods & AI
Computational
Accurate, Scalable and Cross-Platform Cell Identification for High-Resolution Spatial Transcriptomics
Nature Genetics Published 2026-05-20 research article DOI: 10.1038/s41588-026-02610-1
spatial transcriptomics cell segmentation computational method cross-platform imaging single-cell deep learning
Summary: Introduces Cellist, a computationally efficient cell-segmentation method that combines imaging and expression data from high-resolution spatial transcriptomics technologies. As spatial transcriptomics platforms (Visium HD, MERFISH, Xenium, CosMx) achieve subcellular resolution, accurate cell segmentation — delineating individual cell boundaries from spatial expression and imaging data — has become a critical bottleneck. Cellist addresses this by jointly modeling histological image features and transcriptomic spatial patterns to identify cell boundaries, achieving accurate segmentation across multiple commercial spatial transcriptomics platforms without requiring platform-specific parameter tuning. The method is designed for scalability to large tissue sections and demonstrates robust performance across diverse tissue types. By providing accurate, cross-platform cell identification, Cellist enables reliable downstream analyses including cell-type annotation, spatial neighborhood analysis, and cell-cell communication inference that depend on correct cell boundary assignment.
Why it matters: Cell segmentation is the foundational step for all spatial transcriptomics analysis — errors at this stage propagate through every downstream analysis. With the proliferation of high-resolution spatial platforms, a cross-platform segmentation method that works robustly without per-dataset tuning fills a critical gap. Cellist's ability to combine imaging and expression modalities mirrors how human pathologists integrate morphological and molecular information, potentially improving segmentation accuracy beyond either modality alone.
Why for Yiru: Spatial transcriptomics is central to the user's research on the TME. Accurate cell segmentation directly impacts the quality of all downstream analyses — cell-type annotation, spatial neighborhood characterization, and cell-cell communication inference in tumour-immune-stromal contexts. A cross-platform method means consistent analysis pipelines can be applied regardless of which spatial technology generated the data.
Decoding Condition-Specific Cellular Crosstalk in Spatial Omics via Bilinear Edge Classification
bioRxiv (Bioinformatics) Published 2026-05-24 preprint DOI: 10.64898/2026.05.03.722470
spatial transcriptomics cell-cell communication bilinear classification graph neural network condition-specific tissue architecture computational method
Summary: Presents Casei, a bilinear classification framework operating on cellular proximity graphs that directly models condition-specific cell-cell interactions in spatial omics data. While most spatial analysis tools operate at the level of individual cells or cell types, Casei makes cell-cell edges — the interactions between neighboring cells — the fundamental unit of biological inference. The key innovation is a bilinear model whose inductive bias captures coordinated gene-gene relationships between neighboring cells, enabling detection of multicellular interaction patterns that distinguish biological conditions. Applied to liver fibrosis, atherosclerosis, and brain aging, Casei reveals biologically meaningful spatial reorganization: the shift from endothelial- to macrophage-dominated networks in atherosclerotic plaques, disruption of hepatocyte zonation in fibrosis, and oligodendrocyte-microglia crosstalk in aging white matter. Unlike node-centric approaches that may miss coordinated changes between interacting cells, Casei directly tests whether specific cell-type pairs exhibit condition-dependent transcriptional coupling.
Why it matters: Most spatial transcriptomics tools analyze cells independently, yet tissue function emerges from multicellular interactions. Casei's edge-centric perspective — treating cell-cell interactions as the primary analytical unit — represents a conceptual shift that may reveal spatial reorganization invisible to cell-centric methods. The demonstration across fibrosis, atherosclerosis, and brain aging shows broad applicability to diseases where tissue architecture disruption is central to pathology.
Why for Yiru: TME analysis fundamentally depends on understanding how cell-cell interactions change between conditions — tumour versus normal, responder versus non-responder, pre- versus post-treatment. Casei's bilinear edge classification directly addresses this question, making it highly relevant for identifying condition-specific TME reorganization in spatial transcriptomics studies of cancer.
Multiomic State-Transitions Reveal Post-Treatment Transcriptome Desynchronization in Acute Myeloid Leukemia
bioRxiv (Cancer Biology) Published 2026-05-24 preprint DOI: 10.64898/2026.05.20.726707
acute myeloid leukemia state-transition model multiomic transcriptomics miRNA chemotherapy Dlk1-Dio3 temporal dynamics computational biology
Summary: Uses a mathematical state-transition model to study the temporal dynamics of mRNA and miRNA transcriptomes in a mouse model of acute myeloid leukemia (AML) following chemotherapy. The framework represents the mRNA and miRNA transcriptomes as a particle undergoing Brownian motion in a two-dimensional multiomic potential landscape, enabling quantitative characterization of how gene expression programs evolve over time after treatment. A striking finding is the asymmetric post-treatment response: mRNA trajectories respond almost immediately after chemotherapy, whereas miRNA responses are delayed by approximately two weeks — a phenomenon the authors term transcriptome desynchronization. Clustering analysis identifies the imprinted Dlk1-Dio3 locus on chromosome 12qF1 (mouse) / 14q32 (human) as the driver of this temporal delay. While Dlk1-Dio3 has been implicated in acute promyelocytic leukemia and lymphomas, this provides the first evidence linking this locus to AML chemotherapy response and treatment-induced transcriptomic desynchronization. The framework offers a dynamics-based strategy for identifying biological drivers of therapeutic response.
Why it matters: Most transcriptomic studies of drug response capture a single snapshot, missing the temporal dynamics that reveal how different molecular layers respond on different timescales. The discovery that miRNA responses lag behind mRNA by ~2 weeks after chemotherapy has practical implications for clinical monitoring — measuring mRNA alone at early timepoints may miss miRNA-driven resistance programs that emerge later. The Dlk1-Dio3 locus represents a novel candidate biomarker and potential therapeutic target in AML.
Why for Yiru: Temporal modeling of treatment response is relevant beyond AML — the same desynchronization between mRNA and miRNA programs may occur in solid tumours after chemotherapy, targeted therapy, or immunotherapy. Understanding which molecular programs emerge at different timescales after treatment could inform optimal scheduling of combination therapies and timing of response assessment in TME studies.
Zero-Shot De Novo Peptide Sequencing with Open Posttranslational Modification Discovery
Nature Biotechnology Published 2026-05-19 research article DOI: 10.1038/s41587-026-03116-1
de novo peptide sequencing mass spectrometry proteomics post-translational modification deep learning zero-shot computational method
Summary: Introduces RNovA, an open-search de novo peptide sequencing model that achieves zero-shot identification of peptides with unanticipated post-translational modifications (PTMs) directly from tandem mass spectra. Traditional de novo sequencing tools are limited to predicting peptide sequences composed of the 20 standard amino acids, requiring users to specify which PTMs to consider in advance — meaning unknown or unexpected modifications are missed entirely. RNovA overcomes this by modeling the mass spectrum-to-sequence mapping in a way that generalizes to mass shifts not seen during training, enabling discovery of novel PTMs without prior specification. This open-search capability is critical for proteomics studies where the modification landscape is incompletely characterized — including cancer proteomics, where aberrant PTM patterns may drive oncogenic signaling and drug resistance. The model leverages advances in deep learning architecture design to achieve both high sequence accuracy and PTM discovery in a single unified framework.
Why it matters: PTMs regulate virtually every cellular process, yet the full landscape of protein modifications — particularly in disease contexts — remains poorly mapped. RNovA's ability to discover unanticipated PTMs de novo transforms mass spectrometry-based proteomics from a confirmatory technology (testing for known modifications) to a discovery technology (finding new ones). This is particularly valuable for cancer proteomics, where aberrant phosphorylation, acetylation, ubiquitination, and other PTMs drive signaling dysregulation.
Why for Yiru: Proteomics and PTM analysis are increasingly integrated with transcriptomic studies to understand the functional state of the TME. RNovA's ability to discover novel PTMs could reveal modification-driven signaling events in tumour, immune, and stromal compartments that are invisible to transcript-level analysis alone — adding a critical layer of functional annotation to TME multi-omic studies.
AI-Guided Redesign of Laboratory-Evolved Reverse Transcriptases Enhances Prime Editing
Nature Biotechnology Published 2026-05-21 research article DOI: 10.1038/s41587-026-03149-6
prime editing reverse transcriptase protein engineering AI-guided design gene editing directed evolution computational biology
Summary: Reports a computational redesign strategy that improves the efficiency of laboratory-evolved prime editing reverse transcriptases (RTs). Prime editing enables precise genome modifications — insertions, deletions, and base substitutions — without requiring double-strand breaks, but its efficiency is limited by the RT enzyme that synthesizes the edited DNA strand. Previous work used directed evolution to improve prime editing RTs, but the resulting enzymes carried numerous mutations whose individual contributions were unclear. This study applies AI-guided computational redesign to optimized RT variants, identifying which mutations are functionally important and introducing additional beneficial substitutions predicted by structure-based models. The redesigned RTs show enhanced prime editing efficiency across multiple genomic loci and edit types, demonstrating that computational protein design can complement and refine laboratory evolution for genome editing tools.
Why it matters: Prime editing is one of the most versatile precision genome editing technologies, capable of installing all possible base substitutions, small insertions, and small deletions without double-strand breaks. Improving RT efficiency directly expands the therapeutic and research applications of prime editing. The AI-guided redesign approach also demonstrates a generalizable workflow — evolve, then computationally refine — that can be applied to other genome editing enzymes.
Why for Yiru: Genome editing tools are relevant to engineering immune cells for immunotherapy (CAR-T, TCR-T), creating isogenic disease models for TME studies, and developing gene therapies for cancer. Improved prime editing efficiency accelerates the translation of precision genome editing into TME-relevant applications, including engineering T cells with enhanced anti-tumour function and creating precise cancer models.
Genome-Wide Associations of Structural Variants with Human Traits through Imputation from Long-Read Assemblies
Nature Genetics Published 2026-05-20 research article DOI: 10.1038/s41588-026-02612-z
structural variant GWAS long-read sequencing imputation genomics reference panel computational method
Summary: Identifies structural variants (SVs) from long-read genome assemblies, develops a new reference panel and web application to impute SVs from SNP-level genotyping data, and systematically identifies associations between SVs and clinically relevant human traits. SVs — deletions, duplications, inversions, and insertions larger than 50 bp — account for more variable base pairs between individuals than SNPs, yet have been largely invisible to GWAS because they are difficult to genotype at scale. This study bridges that gap by creating a high-quality SV reference panel from long-read assemblies and demonstrating that SVs can be reliably imputed into existing SNP-based cohorts. The resulting SV-GWAS identifies trait-associated SVs that would be missed by SNP-only analysis, revealing new genetic contributors to disease risk and biomarker variation.
Why it matters: The 'missing heritability' problem — where known genetic variants explain only a fraction of disease risk — may partially reflect the invisibility of SVs to standard GWAS. By enabling SV imputation into existing large cohorts, this work democratizes SV analysis without requiring every study to generate long-read data. The identified trait-associated SVs provide new candidates for functional follow-up in disease biology.
Why for Yiru: SVs are increasingly recognized as drivers of cancer — from gene fusions (EWSR1-FLI1) to oncogene amplifications (EGFR, MYC) and tumour suppressor deletions (TP53, PTEN). Methods that enable systematic SV-trait association in large cohorts provide a template for cancer-focused SV studies, including identifying germline SVs that influence tumour risk, progression, and treatment response.
Biomedical discoveries
Biomedicine
Salt-Inducible Kinases as Druggable Targets for Immunotherapy of Ovarian Cancer
Nature Immunology Published 2026-05-20 research article DOI: 10.1038/s41590-026-02512-8
salt-inducible kinase ovarian cancer immunotherapy T cell dysfunction drug repurposing tumour microenvironment immune checkpoint
Summary: Identifies salt-inducible kinases (SIKs) as critical drivers of T cell dysfunction in the immunosuppressive microenvironment of high-grade serous ovarian cancer (HGSOC). Using an all-human high-throughput drug repurposing screen on malignant ascites — the fluid that accumulates in the peritoneal cavity of ovarian cancer patients and contains tumour cells, immune cells, and soluble immunosuppressive factors — the authors discover that SIK inhibition reverses T cell exhaustion and restores anti-tumour effector function. Malignant ascites from HGSOC patients strongly inhibits T cell proliferation, cytokine production, and cytotoxic activity. SIK inhibitors, including existing clinical-grade compounds, rescue T cell function even in the presence of ascites. Mechanistically, SIKs suppress T cell activation by phosphorylating key transcriptional regulators including CRTC2 and class IIa HDACs, keeping them sequestered in the cytoplasm and preventing expression of T cell effector genes. SIK inhibition releases this blockade, reprogramming exhausted T cells toward a functional effector state. In preclinical ovarian cancer models, SIK inhibition synergizes with PD-1 blockade to enhance tumour control.
Why it matters: Ovarian cancer has been largely refractory to checkpoint immunotherapy, with response rates below 15% in unselected patients. The identification of SIKs as a parallel immunosuppressive pathway — operating through a mechanism distinct from PD-1/PD-L1 — opens a new therapeutic axis. Critically, existing SIK inhibitors developed for other indications provide a rapid path to clinical translation. The use of patient-derived malignant ascites for drug screening also establishes a physiologically relevant platform for discovering TME-specific immunotherapy targets.
Why for Yiru: T cell dysfunction in the TME is a central barrier to effective immunotherapy across multiple cancer types. The discovery that SIKs mediate ascites-driven immunosuppression provides a new mechanistic target for reversing T cell exhaustion, with direct relevance to understanding how soluble TME factors — not just cell-surface checkpoints — suppress anti-tumour immunity.
Mapping Where Mouse Models Match — and Miss — Human Tumor Immunity
Nature Immunology Published 2026-05-21 research article DOI: 10.1038/s41590-026-02527-1
tumour microenvironment mouse model human comparative immunology single-cell tumour immunity preclinical models translational research
Summary: Reports a systematic comparison of the tumour microenvironment (TME) between widely used mouse models and human tumours, revealing that these models capture only a limited subset of the cellular diversity and cell-type-specific gene expression programs observed in human cancer. By integrating single-cell RNA-seq data from multiple mouse models (syngeneic transplants, genetically engineered mouse models, and carcinogen-induced tumours) with human tumour atlases across several cancer types, the authors quantify which aspects of tumour immunity are conserved and which are model-specific. Key findings include: (1) mouse tumours consistently show higher proportions of certain immune subsets (e.g., monocytes/neutrophils) and lower proportions of others (e.g., certain T cell exhaustion states) compared to human tumours; (2) gene expression programs within matched cell types show systematic differences, with human T cells expressing more exhaustion markers and mouse macrophages showing distinct activation states; (3) immunotherapy-responsive gene signatures from human trials are only partially recapitulated in mouse models. The study provides a quantitative framework for selecting appropriate mouse models for specific immunotherapy questions and interpreting preclinical results in the context of human tumour immunology.
Why it matters: Mouse models are the backbone of preclinical immunotherapy development, yet the translational failure rate remains high. This systematic comparison provides an evidence-based guide for which models are most appropriate for which questions, and — critically — identifies the specific immune features that are consistently missed by mouse models. This knowledge can prevent overinterpretation of preclinical immunotherapy results and guide the development of improved models that better recapitulate human TME immunology.
Why for Yiru: Understanding the strengths and limitations of preclinical models is essential for designing TME studies that will translate to human cancer. This systematic mouse-human comparison provides a reference for interpreting TME findings from mouse models and identifying which TME features require validation in human samples — directly relevant to computational TME research that spans both species.
p63 and PITX1 Sustain a Pre-Invasive Malignant Keratinocyte Population in Squamous Cell Carcinoma Precursors
bioRxiv (Cancer Biology) Published 2026-05-24 preprint DOI: 10.64898/2026.05.21.725073
squamous cell carcinoma actinic keratosis p63 PITX1 pre-invasive keratinocyte CITE-seq spatial transcriptomics cancer precursor
Summary: Defines the molecular identity of a pre-invasive malignant keratinocyte population in actinic keratosis (AK), the precursor to cutaneous squamous cell carcinoma (cSCC), using single-cell CITE-seq and spatial whole-transcriptome profiling of patient-matched biopsies. The authors identify AK-specific keratinocytes (ASK) — a discrete population localized to the dysplastic basal epidermis, characterized by UV-associated mutational signatures (SBS7b), high mutational burden, and recurrent copy number alterations including 9p loss and 8q gain. ASK cells are maintained in a basal-like undifferentiated state by a ΔNp63/PITX1 regulatory module that simultaneously attenuates Notch/HES1-driven differentiation and activates glycolytic metabolism. Comparison with published invasive cSCC data reveals that ASK share core tumour-propagating gene networks (including IGFBP6, IGFBP2, ITGA6) with tumour-specific keratinocytes but lack invasion effectors (MMP1, MMP10, PTHLH) — suggesting invasion is acquired later. Functional experiments identify IGFBP6 as a pro-proliferative factor in AK-derived cells. The AK microenvironment shows expansion of inflammatory basal keratinocytes, barrier disruption, and early immunosuppressive T cell remodeling, indicating that immune evasion begins at the pre-invasive stage.
Why it matters: cSCC is among the most common human cancers, yet the cellular and molecular transition from pre-invasive actinic keratosis to invasive carcinoma is poorly understood. Identifying ASK as a distinct pre-malignant population governed by p63/PITX1 provides both a conceptual framework for studying the benign-to-malignant transition and candidate targets (ΔNp63, PITX1, IGFBP6) for prevention or interception. The finding that immune remodeling begins at the pre-invasive stage suggests that immunoprevention strategies — targeting the TME before invasion — may be viable.
Why for Yiru: The transition from pre-invasive to invasive cancer involves coordinated changes in both tumour cells and the TME. This study's demonstration that ASK cells share core programs with invasive cancer but lack invasion-specific effectors provides a framework for studying TME remodeling across the invasion continuum — directly relevant to understanding how the TME co-evolves with tumour progression in other cancer types.
A RHNO1-ATR/Chk1 Positive Feedback Loop Sustains the DNA Replication Stress Response
bioRxiv (Cancer Biology) Published 2026-05-24 preprint DOI: 10.64898/2026.05.22.727300
DNA replication stress ATR Chk1 RHNO1 feedback loop genomic instability cancer checkpoint
Summary: Discovers a positive feedback loop wherein the ATR/Chk1 DNA replication stress checkpoint stabilizes the protein RHNO1, which in turn is required to sustain ATR/Chk1 signaling — forming a self-reinforcing circuit that maintains the replication stress response until stress is resolved. While ATR/Chk1 activation is well characterized, how this signal is sustained over the hours required for complete replication stress resolution has been unclear. The authors show that RHNO1 is dispensable for initial ATR/Chk1 activation following replication stress, but is essential for sustained signaling. Under basal conditions, RHNO1 is rapidly degraded by the proteasome; upon replication stress, ATR/Chk1-mediated phosphorylation stabilizes RHNO1 and promotes its localization to stressed replication forks marked by phosphorylated RPA32. RHNO1 depletion does not block initial checkpoint activation but leads to premature checkpoint collapse, accumulation of DNA damage, and genomic instability. RHNO1 knockdown significantly inhibits cancer cell proliferation in vitro and tumour growth in vivo, highlighting its potential as a therapeutic target for tumours reliant on ATR/Chk1 signaling.
Why it matters: Many cancers exhibit elevated replication stress and depend on ATR/Chk1 signaling for survival, making this pathway an attractive therapeutic target. However, ATR inhibitors have shown limited single-agent activity, possibly because cancer cells can maintain residual signaling through feedback mechanisms like the RHNO1 loop. Understanding the distinction between checkpoint activation (which RHNO1 is dispensable for) and checkpoint maintenance (which requires RHNO1) may inform combination strategies — targeting RHNO1 could selectively collapse sustained checkpoint signaling in stressed cancer cells while sparing normal cells that experience only transient stress.
Why for Yiru: DNA damage response and replication stress pathways are relevant to understanding how tumour cells survive genotoxic chemotherapy and how genomic instability shapes tumour evolution. The feedback loop concept — where stress-induced factors sustain the very signaling that stabilizes them — is a general principle that may apply to other stress response pathways in the TME, including hypoxia and metabolic stress responses.
A Glycan-Based Adjuvant Expands the Breadth and Duration of Protection of mRNA-Based Vaccines
Nature Immunology Published 2026-05-22 research article DOI: 10.1038/s41590-026-02517-3
mRNA vaccine adjuvant glycan mannan SARS-CoV-2 germinal center antibody response vaccine technology
Summary: Demonstrates that mannadjuvant — a formulation of fungal mannan and aluminum hydroxide — significantly increases the magnitude, durability, and breadth of the immune response elicited by mRNA-based vaccines. While mRNA vaccines encoding the SARS-CoV-2 spike protein have been remarkably successful, waning antibody titers and reduced efficacy against emerging variants highlight the need for improved durability and cross-variant protection. The mannadjuvant, when co-administered with mRNA-LNP vaccines, enhances germinal center formation, increases the frequency and quality of antigen-specific T follicular helper cells, and broadens the antibody response to recognize diverse viral variants. In mouse models, mannadjuvant-supplemented mRNA vaccines show higher neutralizing antibody titers that persist longer and provide better protection against variant challenge compared to mRNA vaccine alone. The adjuvant works through C-type lectin receptor signaling on dendritic cells, enhancing antigen presentation and innate immune activation.
Why it matters: The rapid waning of mRNA vaccine-induced antibodies and the continuous emergence of new viral variants are major challenges for vaccine durability. A glycan-based adjuvant that enhances germinal center responses addresses both problems simultaneously — stronger and longer-lasting antibody responses with broader variant coverage. Mannan is a well-characterized, inexpensive natural product, making this approach potentially scalable for global vaccine programs. The principles may extend beyond SARS-CoV-2 to mRNA vaccines for other infectious diseases and potentially cancer vaccines.
Why for Yiru: The adjuvant principles discovered here — particularly how C-type lectin receptor signaling shapes the quality and durability of T cell help — are directly relevant to cancer vaccine design. Enhancing germinal center responses and T follicular helper cell quality could improve the efficacy of mRNA-based cancer vaccines targeting neoantigens or tumour-associated antigens in the TME context.
Branched-Chain Keto Acids Secreted by Cancer Cells Suppress Anti-Tumour Immunity via NOTCH
Nature Immunology Published 2026-05-22 research article DOI: 10.1038/s41590-026-02528-0
tumour microenvironment metabolism branched-chain keto acid NOTCH T cell immunosuppression cancer metabolism immunometabolism
Summary: Reveals that cancer cells suppress anti-tumour immunity through secretion of branched-chain keto acids (BCKAs) into the tumour microenvironment, where they are taken up by tumour-associated macrophages (TAMs) and signal through NOTCH to drive immunosuppressive polarization. BCKAs are metabolic intermediates of branched-chain amino acid (BCAA) catabolism, a pathway frequently upregulated in cancer cells to support proliferation. The study shows that BCKAs — particularly α-ketoisocaproate (KIC) — act not merely as metabolic fuel but as intercellular signaling molecules. TAMs that take up tumour-derived BCKAs activate NOTCH signaling, which drives expression of immunosuppressive factors including IL-10 and TGF-β while suppressing expression of pro-inflammatory cytokines. Genetic or pharmacological disruption of BCKA secretion by tumour cells, or blockade of NOTCH signaling in TAMs, restores anti-tumour T cell responses and enhances the efficacy of checkpoint immunotherapy in preclinical models. The findings establish a direct mechanistic link between cancer cell metabolism and immune suppression mediated through secreted metabolites.
Why it matters: The metabolic interplay between tumour cells and immune cells in the TME is increasingly recognized as a determinant of immunotherapy response. BCKAs represent a new class of tumour-derived immunosuppressive metabolites that function through specific receptor-mediated signaling (NOTCH) rather than simply through nutrient competition. This distinguishes them from better-known mechanisms like glucose depletion or lactate-mediated suppression and opens new therapeutic opportunities — targeting BCKA production, secretion, or sensing could reverse metabolic immunosuppression without the broad toxicity of global metabolic inhibitors.
Why for Yiru: Immunometabolism — how metabolic crosstalk between tumour cells and immune cells shapes anti-tumour immunity — is directly relevant to understanding TME heterogeneity and identifying metabolic vulnerabilities. The BCKA-NOTCH axis provides a concrete molecular mechanism connecting a common cancer metabolic pathway (BCAA catabolism) to immune suppression, which can be analyzed in spatial TME studies combining metabolomic and transcriptomic readouts.
Cross-disciplinary watchlist
Other Fields
Allele-Specific Methylation Uncovers Non-Mendelian Inheritance of Epigenetic Patterns Across Generations
Nature Genetics Published 2026-05-20 research article DOI: 10.1038/s41588-026-02604-z
epigenetics DNA methylation non-Mendelian inheritance allele-specific long-read sequencing mouse genetics imprinting
Summary: Investigates allele-specific DNA methylation inheritance patterns in mouse liver and muscle using a genome-wide framework based on long-read sequencing. While DNA methylation is generally considered to be reset during embryogenesis, emerging evidence suggests that certain methylation patterns can escape reprogramming and be transmitted across generations. The authors develop an allele-specific methylation profiling approach using Oxford Nanopore long-read sequencing that distinguishes methylation on maternally versus paternally inherited alleles, enabling systematic detection of parent-of-origin effects. Across the genome, most methylation patterns follow Mendelian inheritance, but approximately 7% show non-Mendelian behavior — including discovery of new imprinted genes and a striking paramutation-like effect at the Capn11 locus where one allele's methylation state influences the other. The companion News & Views article frames this as revealing 'new imprinted genes and a paramutation at the Capn11 locus,' highlighting the paradigm-shifting nature of finding paramutation-like phenomena in mammals.
Why it matters: Non-Mendelian epigenetic inheritance challenges the fundamental assumption that epigenetic marks are completely reset between generations. The discovery that ~7% of methylation patterns show non-Mendelian inheritance — including mammalian paramutation — suggests that environmental exposures or parental experiences could leave transgenerational epigenetic marks. This has profound implications for understanding disease risk, particularly for conditions with familial clustering but no identified genetic cause.
Why for Yiru: Epigenetic dysregulation is a hallmark of cancer, and understanding the rules of epigenetic inheritance — including non-Mendelian patterns — may illuminate why some individuals have elevated cancer risk beyond what genetics predicts. Long-read-based allele-specific methylation analysis could be applied to tumour-normal paired samples to identify cancer-associated epigenetic inheritance patterns.
Patterns and Drivers of 43,617 Mosaic Chromosomal Alterations in Blood
Nature Genetics Published 2026-05-19 research article DOI: 10.1038/s41588-026-02592-0
mosaic chromosomal alteration clonal hematopoiesis UK Biobank whole-genome sequencing somatic mutation aging blood genomics
Summary: Reports high-resolution analyses of blood-derived whole-genome sequence data from UK Biobank, detecting and characterizing 43,617 mosaic chromosomal alterations (mCAs) — large somatic copy-number changes present in a subset of blood cells. mCAs are hallmarks of clonal hematopoiesis, an age-related phenomenon where hematopoietic stem cells carrying somatic mutations expand to generate a detectable fraction of blood cells. Using deep whole-genome sequencing, the authors achieve substantially higher sensitivity than previous array-based studies, detecting new mCAs including copy-neutral loss of heterozygosity events and small subclonal alterations. The study identifies rare germline protein-coding variants associated with increased risk of developing specific mCAs, implicating genes involved in DNA damage repair and telomere maintenance as determinants of clonal expansion. Associations between specific mCAs and risk of hematologic malignancies, cardiovascular disease, and all-cause mortality are refined with the increased statistical power of this large cohort.
Why it matters: Clonal hematopoiesis is one of the most important discoveries in aging biology — it represents a premalignant state that also independently predicts cardiovascular and all-cause mortality. This study more than doubles the resolution of mCA detection and identifies the germline genetic variants that predispose individuals to clonal expansion, providing both biomarkers for risk stratification and mechanistic insights into how clonal hematopoiesis arises. The scale (43,617 mCAs) makes this the definitive resource for understanding somatic chromosomal evolution in blood.
Why for Yiru: Clonal hematopoiesis is relevant to cancer biology in multiple ways: it represents the earliest detectable step toward hematologic malignancy, it influences the immune system composition (since mutant clones can differentiate into immune cells), and it may affect responses to chemotherapy and immunotherapy. Understanding the genetic determinants of clonal expansion provides a framework for studying somatic evolution in other tissues.
Resetting Autoimmune Disease with CAR Cell Therapies
Nature Medicine Published 2026-05-21 review DOI: 10.1038/s41591-026-04430-6
CAR T cell autoimmune disease cell therapy B cell depletion immune reset clinical translation review
Summary: Surveys the rapidly developing landscape of CAR T cell therapies for autoimmune disease from both technological and clinical standpoints. Originally developed for B cell malignancies, CAR T cells targeting CD19 or BCMA achieve deep B cell depletion that can 'reset' the immune system — eliminating autoreactive B cells and plasma cells while allowing regeneration of a naive B cell repertoire. Early clinical results in systemic lupus erythematosus, systemic sclerosis, and myositis have shown remarkable efficacy, with some patients achieving drug-free remission. The review covers the concept of deep B cell depletion as immune reset, compares different CAR constructs and target antigens (CD19 vs. BCMA vs. dual-targeting), discusses safety considerations including cytokine release syndrome and long-term B cell aplasia, and explores the potential expansion to T-cell-targeted CAR therapies for T-cell-mediated autoimmune diseases. The authors frame CAR T therapy as potentially curative for autoimmune disease — analogous to its transformative impact in hematologic malignancies.
Why it matters: The expansion of CAR T therapy from oncology to autoimmune disease represents one of the most exciting translational developments in immunology. If the early promising results are confirmed in larger trials, CAR T cells could fundamentally change the treatment paradigm for severe autoimmune diseases — moving from chronic immunosuppression to one-time immune reset. The technological lessons from oncology CAR T development (construct optimization, toxicity management, manufacturing) are directly applicable, accelerating clinical translation.
Why for Yiru: CAR T cell therapy is a platform technology whose principles — engineering immune cells to recognize and eliminate target cells — span oncology and autoimmunity. The concept of immune reset through deep B cell depletion is directly relevant to understanding how the immune system reconstitutes after therapy, which is also important for interpreting TME immune dynamics after lymphodepleting chemotherapy or CAR T therapy in cancer.
MOSAIC: A Multimodal Adaptive Optical Microscope for In Vivo Imaging from Molecules to Organisms
Nature Methods Published 2026-05-22 research article DOI: 10.1038/s41592-026-03066-1
microscopy multimodal imaging light-sheet two-photon in vivo imaging adaptive optics instrumentation
Summary: Introduces MOSAIC (Multimodal Optical System for Adaptive Imaging across scales), a versatile microscope platform that enables seamless transition between light-sheet microscopy, two-photon microscopy, label-free imaging modalities, and super-resolution techniques within a single instrument. The key innovation is an adaptive optical design that reconfigures the illumination and detection paths to optimize for each imaging modality without requiring manual realignment. MOSAIC can image samples spanning spatial scales from single molecules to whole organisms, and temporal scales from milliseconds to days. The platform demonstrates applications including: whole-brain activity imaging in larval zebrafish with light-sheet microscopy, deep tissue imaging of immune cell dynamics in mouse lymph nodes with two-photon microscopy, and label-free metabolic imaging using endogenous fluorescence. By integrating multiple modalities into a single reconfigurable platform, MOSAIC eliminates the need for separate specialized microscopes and enables correlative imaging workflows where the same sample is imaged with multiple modalities.
Why it matters: Biological questions increasingly require bridging spatial and temporal scales — from molecular dynamics to tissue-level organization. MOSAIC's reconfigurable design addresses a practical bottleneck: most labs cannot afford or accommodate multiple specialized microscope systems. By integrating modalities that are typically on separate instruments, MOSAIC enables experiments that are currently impractical, such as following immune cell migration with two-photon microscopy and then capturing the whole-organ context with light-sheet imaging.
Why for Yiru: Advanced microscopy is essential for TME research — imaging immune cell infiltration, tumour-stromal interactions, and drug distribution in intact tissues. MOSAIC's ability to combine deep-tissue two-photon imaging (for tracking immune cells in live tumours) with whole-tissue light-sheet imaging (for comprehensive spatial context) is directly relevant to building multi-scale spatial maps of the TME.
The Combination of EWSR1-FLI1 and Loss of One EWSR1 Allele Leads to the Induction of Trisomy 8
bioRxiv (Cancer Biology) Published 2026-05-24 preprint DOI: 10.64898/2026.05.21.726567
Ewing sarcoma EWSR1-FLI1 trisomy 8 chromosomal instability Aurora B pediatric cancer fusion oncogene
Summary: Elucidates the mechanism by which trisomy 8 — the most common secondary chromosomal aberration in Ewing sarcoma, associated with poor prognosis — is induced. Using a conditional cell line system that inducibly expresses the EWSR1-FLI1 fusion oncogene while simultaneously degrading the remaining wild-type EWSR1 allele (mimicking the genetic configuration of Ewing sarcoma cells), the authors show that the combination of EWSR1-FLI1 expression and EWSR1 haploinsufficiency induces a high incidence of trisomy 8 within eight days. Mechanistically, EWSR1 knockout alone is sufficient to cause trisomy 8, which can be rescued by wild-type EWSR1 but not by an EWSR1 mutant (R565A) that cannot interact with Aurora B kinase — implicating the EWSR1-Aurora B pathway in chromosome segregation fidelity. The findings establish a direct causal link between the defining genetic event of Ewing sarcoma (EWSR1-FLI1 formation with concomitant EWSR1 loss) and the most common secondary genetic abnormality (trisomy 8), providing a mechanistic explanation for the karyotypic evolution of this tumour.
Why it matters: Secondary genetic changes in cancer — such as trisomy 8 in Ewing sarcoma — are typically assumed to arise stochastically from general genomic instability, with selection favoring clones that happen to carry advantageous alterations. This study shows instead that the initiating oncogenic event directly induces a specific secondary abnormality, demonstrating a deterministic path of karyotypic evolution. The EWSR1-Aurora B connection suggests that therapeutic targeting of Aurora B kinase could prevent the emergence of trisomy 8 clones and potentially improve outcomes in Ewing sarcoma.
Why for Yiru: Understanding how oncogenic fusions drive specific patterns of aneuploidy is relevant to the broader question of how tumour genotype shapes chromosomal evolution. The concept that an initiating oncogene can deterministically induce specific secondary genetic changes — rather than simply creating permissive conditions for random alterations — has implications for predicting tumour evolution and identifying early intervention points.
enCORE: Network-Based Prediction of Clustered Open Regulatory Elements from Single-Cell Chromatin Accessibility
bioRxiv (Bioinformatics) Published 2026-05-23 preprint DOI: 10.64898/2026.03.17.712366
single-cell ATAC-seq regulatory element enhancer chromatin accessibility gene regulation network biology epigenomics
Summary: Introduces enCORE, a computational framework that identifies Clustered Open Regulatory Elements (COREs) — higher-order cis-regulatory structures spanning tens to hundreds of kilobases — from single-cell ATAC-seq data by leveraging enhancer-enhancer interaction networks. Most single-cell chromatin accessibility tools analyze individual peaks independently, missing the coordinated, domain-level regulation essential for cell-type-specific gene expression. enCORE models the co-accessibility relationships between open chromatin regions to assemble COREs, recovering known enhancer clusters and identifying cell-type-specific regulatory domains. Applied to hematopoietic differentiation, enCORE recapitulates established lineage hierarchies and resolves lineage-specific programs enriched for canonical master transcription factors and fine-mapped autoimmune disease GWAS variants. In colorectal cancer, enCORE captures tumour-associated H3K27ac landscapes and prioritizes USP7 as a potential therapeutic candidate through in silico perturbation analysis. The framework provides a scalable platform for moving beyond peak-level analysis to domain-level epigenetic interpretation.
Why it matters: Gene regulation operates at the level of enhancer clusters and regulatory domains, not individual enhancers. enCORE fills a critical gap in the single-cell epigenomics toolkit by identifying these higher-order structures directly from single-cell ATAC-seq data — without requiring Hi-C or other 3D genome data. The ability to detect regulatory domain reorganization in cancer and prioritize potential therapeutic targets through in silico perturbation demonstrates the translational potential of domain-level epigenomic analysis.
Why for Yiru: Regulatory element analysis in the TME — understanding how enhancer landscapes differ between tumour cells, immune cells, and stromal cells — requires tools that capture domain-level regulation. enCORE's ability to identify disease-associated regulatory domains and prioritize targets from single-cell epigenomic data is directly relevant to studying how chromatin-level dysregulation shapes TME cell states in cancer.