Research Radar — 2026-05-06
Methods & AI
Computational
TxPert: using multiple knowledge graphs for prediction of transcriptomic perturbation effects
Nature Biotechnology Published 2026-05-01 research article DOI: 10.1038/s41587-026-03113-4
knowledge graph transcriptomic perturbation deep learning drug discovery systems biology
Summary: TxPert integrates multiple knowledge graphs with deep learning to predict transcriptomic perturbation effects, enabling systematic exploration of how genetic interventions alter cellular states without exhaustive experimental screening.
Why it matters: Predicting perturbation outcomes from knowledge graphs could dramatically reduce the experimental burden of mapping gene function — a core challenge in functional genomics and drug target identification.
Why for Yiru: Knowledge graph-based perturbation modeling has direct crossover to predicting immune cell responses to CAR-T engineering and checkpoint perturbations in the tumor microenvironment.
Autonomous pathology research using agentic AI shows potential in oncology
Nature Medicine Published 2026-05-05 research article DOI: 10.1038/s41591-026-04403-9
agentic AI digital pathology oncology biomedical AI diagnostic AI
Summary: The agentic AI tool SPARK autonomously reproduces pathology-based reasoning, generating biological hypotheses and deriving diagnostic, prognostic, and predictive cellular parameters from histopathology data without human guidance.
Why it matters: This represents one of the first demonstrations of autonomous scientific reasoning by AI in a biomedical domain — suggesting a future where AI agents can independently conduct pathology research workflows.
Why for Yiru: Agentic AI for biomedical reasoning is directly relevant to computational immunology — autonomous analysis of tumor microenvironment features could accelerate spatial omics interpretation.
MicroSplit: semantic unmixing of fluorescent microscopy data
Nature Methods Published 2026-05-05 research article DOI: 10.1038/s41592-026-03082-1
computational microscopy semantic unmixing deep learning image analysis fluorescence imaging
Summary: MicroSplit is a computational method for semantic unmixing of fluorescence microscopy data, disentangling overlapping fluorescent signals into their constituent biological components to overcome optical limits and improve multiplexed imaging analysis.
Why it matters: Semantic unmixing addresses a fundamental limitation in multiplexed fluorescence imaging — the inability to cleanly separate overlapping signals — which is critical for spatial biology applications.
Why for Yiru: Image analysis methods that improve multiplexed fluorescence data quality are directly applicable to spatial proteomics and spatial transcriptomics workflows in tumor microenvironment studies.
Digital twins of ex vivo human lungs enable accurate and personalized evaluation of therapeutic efficacy
Nature Biotechnology Published 2026-05-04 research article DOI: 10.1038/s41587-026-03121-4
digital twin computational model organ-level modeling personalized medicine therapeutic evaluation
Summary: A comprehensive digital twin of ex vivo human lungs integrating molecular, physiological, functional, and clinical data enables accurate and personalized evaluation of therapeutic efficacy through computational simulation.
Why it matters: Organ-level digital twins represent a paradigm shift for precision medicine — enabling in silico therapeutic evaluation that could transform drug development and clinical decision-making.
Why for Yiru: Digital twin AI for organs directly aligns with translational medicine goals — organ-level computational models could integrate spatial omics data and immune microenvironment simulations for personalized immunotherapy.
Single-cell data integration across weakly linked modalities
PLOS Computational Biology Published 2026-05-06 research article DOI: 10.1371/journal.pcbi.1014231
single-cell multi-omics integration data integration bioinformatics weakly linked modalities
Summary: A computational method for integrating single-cell data across weakly linked modalities — where correlations between data types are tenuous — addressing the challenge of emerging measurement technologies that produce sparsely correlated multimodal readouts.
Why it matters: As new single-cell modalities emerge with weak correlations to existing data types, robust integration methods are essential to unlock the full potential of multi-omics atlases.
Why for Yiru: Single-cell multi-omics integration is central to computational immunology — methods for weak-linkage integration could improve mapping between scRNA-seq, scATAC-seq, and spatial data in tumor microenvironment studies.
Biomedical discoveries
Biomedicine
Distinct in vivo dynamics of donor-derived stem cell memory CAR T cells post-allogeneic HSCT relapse
Cell Published 2026-04-28 research article DOI: 10.1016/j.cell.2026.03.047
CAR-T stem cell memory allogeneic HSCT immunotherapy T cell dynamics
Summary: In vivo tracking reveals distinct engraftment and dynamics of donor-derived stem cell memory CAR T cells following allogeneic hematopoietic stem cell transplantation relapse, identifying key determinants of CAR-T persistence and anti-tumor function in the post-transplant setting.
Why it matters: Understanding CAR-T cell dynamics after allogeneic transplant could improve the design of cellular immunotherapies for hematologic malignancies, where relapse remains a major clinical challenge.
Why for Yiru: CAR-T cell biology, stem cell memory T cell persistence, and in vivo dynamics in the post-transplant tumor microenvironment are directly aligned with Boss's core interests in immunotherapy and T cell biology.
Tissue tension fosters macrophage-driven lipid peroxidation-induced DNA damage
Cancer Cell Published 2026-04-28 research article DOI: 10.1016/j.ccell.2026.03.022
tumor microenvironment macrophage tissue mechanics DNA damage lipid peroxidation
Summary: Tissue-level mechanical tension in the tumor microenvironment drives macrophages to produce lipid peroxidation that induces DNA damage, revealing a mechano-immunological mechanism of mutagenesis that links physical tissue properties to genomic instability.
Why it matters: This discovery bridges tissue mechanics and cancer immunology — showing that physical forces in the TME can directly drive mutagenesis through immune cell mediators, opening new angles for cancer prevention.
Why for Yiru: Macrophage biology in the tumor microenvironment, mechano-immunology, and DNA damage mechanisms are all central to Boss's research interests — this paper connects tissue tension to immune-driven mutagenesis.
Activated T cell extracellular vesicle DNA transfer enhances antigen presentation and anti-tumor immunity
Cancer Cell Published 2026-04-28 research article DOI: 10.1016/j.ccell.2026.03.023
T cell extracellular vesicle antigen presentation anti-tumor immunity DNA transfer
Summary: Activated T cells transfer DNA via extracellular vesicles that enter the nucleus of recipient cells, enhancing antigen presentation and boosting anti-tumor immunity through a previously unappreciated intercellular communication mechanism.
Why it matters: The discovery that T cell-derived EV DNA enters recipient cell nuclei to enhance antigen presentation reveals a novel intercellular communication axis that could be harnessed for cancer immunotherapy.
Why for Yiru: T cell biology, anti-tumor immunity, and intercellular communication in the tumor microenvironment — this finding has implications for understanding CAR-T mechanisms and designing next-generation immunotherapies.
High-resolution single-cell mapping of clonal hematopoiesis and structural variation in aplastic anemia
Nature Genetics Published 2026-05-01 research article DOI: 10.1038/s41588-026-02587-x
single-cell clonal hematopoiesis structural variation aplastic anemia hematopoietic stem cells
Summary: High-resolution single-cell mapping of aplastic anemia reveals the clonal architecture of hematopoiesis driven by HLA loss and structural variation, providing a detailed view of how T-cell-mediated immune destruction reshapes the hematopoietic landscape.
Why it matters: Single-cell resolution of clonal hematopoiesis in the context of immune-mediated marrow failure provides mechanistic insight into how immune selection shapes clonal evolution — relevant to understanding pre-malignant states.
Why for Yiru: Single-cell genomics of hematopoiesis with immune-mediated selection pressure connects to computational immunology interests in clonal dynamics and immune surveillance in the bone marrow microenvironment.
Spatially decoding genotype-associated epigenetic landscapes in human lymphoma FFPE tissues via epi-Patho-DBiT
Nature Communications Published 2026-05-01 research article DOI: 10.1038/s41467-026-71576-9
spatial epigenomics FFPE lymphoma chromatin accessibility spatial biology
Summary: Epi-Patho-DBiT enables spatially resolved co-profiling of chromatin accessibility and whole transcriptome in FFPE human lymphoma tissues, decoding how genotype-associated epigenetic landscapes vary across tissue regions in B-cell lymphomas.
Why it matters: Spatial epigenomic profiling in FFPE tissues — the standard clinical archive format — dramatically expands the translational reach of spatial biology technologies to retrospective clinical cohorts.
Why for Yiru: Spatial multi-omics in clinical lymphoma FFPE samples is directly relevant to Boss's interests in spatial transcriptomics and tumor microenvironment analysis — this method bridges genomics, epigenomics, and spatial context.
Disrupted molecular glue complex drives RAS inhibitor resistance
Cell Published 2026-05-05 research article DOI: 10.1016/j.cell.2026.03.031
RAS inhibitor drug resistance molecular glue cancer biology targeted therapy
Summary: Disruption of a molecular glue complex that mediates RAS degradation is identified as a mechanism driving resistance to RAS inhibitors, revealing a new mode of acquired drug resistance in RAS-driven cancers.
Why it matters: Understanding resistance mechanisms to RAS inhibitors — long considered undruggable — is critical as these therapies enter clinical use; this discovery identifies a resistance pathway that could be therapeutically targeted.
Why for Yiru: Cancer drug resistance mechanisms are relevant to understanding treatment failure in the tumor microenvironment context and could inform combination strategies with immunotherapy.
Cross-disciplinary watchlist
Other Fields
Force-free molecular dynamics through autoregressive equivariant networks
Nature Machine Intelligence Published 2026-05-05 research article DOI: 10.1038/s42256-026-01227-7
molecular dynamics equivariant network autoregressive model AI for science computational physics
Summary: An autoregressive equivariant neural network performs force-free molecular dynamics simulations without explicit force computation, overcoming timescale and system-size limitations of traditional MD while maintaining physical consistency through symmetry-preserving architecture.
Why it matters: Removing the force computation bottleneck in MD simulations could enable exploration of biological processes at timescales previously inaccessible — with implications for protein dynamics and drug binding studies.
Why for Yiru: Equivariant networks for molecular simulation represent state-of-the-art AI-for-science methodology — the symmetry-preserving approach could inspire new architectures for modeling immune receptor-ligand interactions.
Adaptive spatial hashing with dual-domain memristive hardware
Nature Communications Published 2026-05-02 research article DOI: 10.1038/s41467-026-72743-8
memristive computing approximate search AI hardware locality-sensitive hashing energy-efficient AI
Summary: A dual-domain memristive hardware architecture implements adaptive locality-sensitive hashing for energy-efficient approximate similarity search, overcoming fixed-threshold limitations and analog encoding inefficiencies of prior approaches.
Why it matters: Energy-efficient hardware for similarity search could dramatically reduce the computational cost of large-scale AI retrieval and database operations — a critical bottleneck as AI models scale.
Why for Yiru: Novel AI hardware architectures that enable efficient large-scale similarity search could eventually accelerate computational biology applications like single-cell data retrieval and molecular similarity searches.
Swimming with robots: investigating fish locomotion, sensing, and schooling behavior with robotic swimmers
Nature Communications Published 2026-05-02 review article DOI: 10.1038/s41467-026-72478-6
robotics bio-inspired AI locomotion swarm intelligence sensorimotor control
Summary: A comprehensive review of how robotic fish systems have been used to investigate and model fish locomotion, sensing, schooling, and collective behavior — demonstrating bidirectional knowledge transfer between robotics and biology.
Why it matters: Bio-inspired robotics that reveal biological principles through physical modeling represents an emerging paradigm — robotic systems serve as testable hypotheses about how organisms achieve complex behaviors.
Why for Yiru: The robotics-biology interface demonstrates how AI and physical modeling can decode complex biological systems — a principle applicable to modeling immune cell behavior and cell-cell interactions in tissue contexts.