Research Radar — 2026-05-17
Methods & AI
Computational
De novo design of peptides localizing at the interface of biomolecular condensates
Nature Communications Published 2026-05-16 research article DOI: 10.1038/s41467-026-73099-9
de novo protein design biomolecular condensates phase separation peptide design Rosetta
Summary: Presents a computational pipeline for de novo design of peptides that specifically localize to the interface of biomolecular condensates. Using structure-based design principles, the authors create peptides that partition to condensate surfaces rather than interiors, enabling programmable targeting of phase-separated compartments without disrupting their internal organization. Validated across multiple condensate systems in vitro and in cells.
Why it matters: Programmable targeting of condensate interfaces opens a new dimension in manipulating phase-separated biology. Rather than dissolving or hardening condensates, interface-localizing peptides could modulate condensate communication, cargo exchange, and material properties — with therapeutic implications for condensate-linked diseases including neurodegeneration and cancer.
Why for Yiru: Protein design and condensate biology intersect at a frontier relevant to understanding biomolecular organization in the TME. The ability to computationally design peptides that target specific subcellular compartments connects to spatial biology and targeted delivery interests.
HESpotEx: a dual-stream deep learning framework for spot-level gene expression prediction from histological images
Nature Computational Science Published 2026-05-15 research article DOI: 10.1038/s43588-026-00992-0
spatial transcriptomics deep learning histology gene expression prediction computational pathology
Summary: Introduces HESpotEx, a dual-stream deep learning framework that predicts spot-level gene expression directly from H&E-stained histological images. The architecture separately processes tissue morphology and spatial context streams before fusing them for expression prediction. Demonstrates superior performance over existing methods across multiple tissue types and spatial transcriptomics platforms, effectively capturing spatially variable gene expression patterns from routine histology.
Why it matters: Predicting gene expression from routine histology images could democratize spatial transcriptomics by eliminating the need for costly sequencing. If validated clinically, this approach could bring spatially-resolved molecular profiling to every pathology workflow — transforming diagnostic and research capabilities.
Why for Yiru: Spatial transcriptomics and deep learning integration is directly aligned with Boss's research interests in tumor microenvironment spatial biology. Methods that extract molecular information from histological images bridge the gap between routine pathology and high-dimensional molecular profiling.
Aberration-aware 3D localization microscopy via self-supervised neural-physics learning
Nature Communications Published 2026-05-16 research article DOI: 10.1038/s41467-026-73045-9
super-resolution microscopy self-supervised learning computational imaging 3D localization neural-physics
Summary: Develops a self-supervised neural-physics framework for aberration-aware 3D localization microscopy. The method learns to correct optical aberrations without requiring ground-truth calibration data by integrating physical models of light propagation with neural network training. Achieves high-precision 3D single-molecule localization even under severe optical aberrations that would degrade conventional methods.
Why it matters: Optical aberrations are a universal challenge in microscopy, especially for deep-tissue imaging. A self-supervised solution that does not require calibration samples or aberration measurements could make high-quality 3D super-resolution accessible to any lab with a standard microscope.
Why for Yiru: Advanced microscopy and computational imaging are essential for spatial biology. Methods that improve imaging quality in challenging tissue environments directly support high-resolution TME characterization.
Relational biological structure improves fine-mapping of causal GWAS variants under weak signal
bioRxiv Published 2026-05-16 preprint DOI: 10.1101/2026.05.15.725513
GWAS fine-mapping statistical genetics biological structure weak signal
Summary: Proposes a method that incorporates relational biological structure — such as gene regulatory networks, protein-protein interactions, and pathway annotations — to improve fine-mapping of causal GWAS variants, particularly under weak association signals. Demonstrates that leveraging biological prior knowledge substantially increases power to resolve causal variants in regions where standard statistical fine-mapping fails.
Why it matters: Most GWAS loci contain multiple correlated variants with weak individual effects, making causal variant identification notoriously difficult. Incorporating biological structure as a prior is a principled way to boost fine-mapping resolution without requiring larger sample sizes — directly addressing a major bottleneck in translating GWAS findings to biological mechanisms.
Why for Yiru: Understanding the genetic basis of complex traits including cancer susceptibility and immune phenotypes requires robust fine-mapping. Methods that integrate biological knowledge with statistical genetics align with interests in connecting genomics to disease mechanisms.
Talk2QSP: Deriving Executable Scenarios from Unstructured Literature via Human-in-the-Loop Agents
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2026.05.06.723244
quantitative systems pharmacology AI agents literature mining human-in-the-loop QSP modeling
Summary: Introduces Talk2QSP, an AI agent framework that derives executable quantitative systems pharmacology (QSP) models from unstructured biomedical literature through human-in-the-loop interaction. The system reads papers, extracts mechanistic relationships, formulates mathematical models, and allows researchers to iteratively refine model structure and parameters through natural language dialogue. Validated on immunotherapy and oncology case studies.
Why it matters: Building QSP models currently requires months of expert manual curation. An AI agent that can draft executable models from literature and engage researchers in refining them could dramatically accelerate model-driven drug development and reduce the barrier to entry for QSP approaches.
Why for Yiru: AI-driven extraction of mechanistic models from literature connects to interests in systems biology and computational approaches to understanding complex biological systems including the TME. The human-in-the-loop paradigm is also relevant to how AI assistants can augment research workflows.
Large-scale discovery, analysis and design of protein energy landscapes
Nature Published 2026-05-13 research article DOI: 10.1038/s41586-026-10465-z
protein energy landscapes computational biophysics protein design molecular dynamics structural biology
Summary: Reports a large-scale computational framework for discovering, analyzing, and designing protein energy landscapes at unprecedented scale. Maps the conformational energy landscapes of thousands of proteins, revealing general principles governing folding, allostery, and functional dynamics. Demonstrates that designed energy landscapes can program specific functional behaviors into proteins.
Why it matters: Protein energy landscapes govern every aspect of protein function — folding, binding, catalysis, and allostery. A systematic framework for mapping and designing these landscapes at scale represents a fundamental advance in protein science, with direct applications to enzyme engineering and therapeutic protein design.
Why for Yiru: Protein biophysics and design are foundational to understanding molecular recognition in biological systems. Principles of energy landscape design have conceptual parallels to understanding cellular state landscapes in development and disease.
Biomedical discoveries
Biomedicine
Ecotypes of triple-negative breast cancer in response to chemotherapy
Nature Published 2026-05-13 research article DOI: 10.1038/s41586-026-10469-9
triple-negative breast cancer ecotypes chemotherapy tumor microenvironment transcriptional archetypes
Summary: Defines transcriptional ecotypes of triple-negative breast cancer (TNBC) and characterizes their differential responses to chemotherapy. Using multi-omics profiling of a large TNBC cohort, the study identifies distinct tumor-immune ecosystem archetypes that predict chemotherapy outcomes independently of standard clinical factors. Certain ecotypes show intrinsic chemoresistance driven by specific stromal-immune interactions.
Why it matters: TNBC is the most aggressive breast cancer subtype with highly variable chemotherapy responses. Ecotype-based stratification could guide treatment decisions — identifying patients likely to benefit from standard chemotherapy versus those needing alternative approaches — and reveals targetable microenvironmental mechanisms of chemoresistance.
Why for Yiru: Tumor ecotypes and microenvironment-driven therapy responses are at the core of Boss's research interests. Understanding how the TME composition shapes chemotherapy outcomes in breast cancer directly informs spatial biology and immunotherapy research directions.
Glutamatergic neuron-tumor synapses shape human glioblastoma cell states through radial glia plasticity
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2026.05.14.725216
glioblastoma neuron-tumor synapse glutamatergic signaling radial glia tumor plasticity
Summary: Reveals that glutamatergic neuron-tumor synapses drive human glioblastoma cell state transitions through activation of a radial glia developmental program. Using single-cell and spatial transcriptomics of patient tumors, the authors show that synaptic input from surrounding neurons pushes GBM cells into a proliferative, stem-like state resembling outer radial glia — a cell type critical for human cortical development. Pharmacological blockade of glutamatergic signaling reduces tumor growth in patient-derived models.
Why it matters: The discovery that glioblastoma cells hijack a developmental neuron-glia signaling program to maintain stemness and drive growth represents a new mechanistic axis in brain cancer biology. It positions glutamatergic neurotransmission as a therapeutic vulnerability distinct from conventional oncogenic targets.
Why for Yiru: The intersection of neuroscience and cancer biology — particularly how neural activity shapes tumor cell states — is an emerging frontier. The single-cell and spatial approaches used here are methodologically relevant to TME research.
Beyond RECIST: mathematical modeling and Bayesian inference reveal the importance of immune parameters in metastatic breast cancer
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2025.08.11.669777
mathematical oncology breast cancer immunotherapy Bayesian inference RECIST tumor-immune dynamics
Summary: Fits mathematical models of tumor-immune dynamics to metastatic breast cancer patients receiving checkpoint inhibition combined with entinostat, integrating RECIST measurements and spatial proteomics. Bayesian parameter inference reveals that only immune-suppressive parameters control therapeutic response — parameters governing immune cytotoxicity are uninformative. Suggests that overcoming immunosuppression, rather than boosting killing capacity, is the rate-limiting step in immunotherapy response.
Why it matters: The finding that immunosuppression — not immune activation or cytotoxicity — controls response has profound implications for immunotherapy design. It argues for prioritizing strategies that relieve suppression (e.g., Treg depletion, myeloid reprogramming) over those that enhance effector function (e.g., co-stimulation).
Why for Yiru: Mechanistic modeling of tumor-immune dynamics combined with spatial proteomics is directly relevant to understanding TME biology quantitatively. The conclusion that immunosuppression is the dominant parameter aligns with interests in immune evasion mechanisms.
Circadian Clock Programming of Anticipatory Antiviral Immunity Gates Enteric Virus Infection Susceptibility
bioRxiv Published 2026-05-16 preprint DOI: 10.1101/2026.05.15.725500
circadian rhythm antiviral immunity enteric virus interferon chrono-immunology
Summary: Demonstrates that the circadian clock programs anticipatory antiviral immune defenses in the intestinal epithelium, creating time-of-day-dependent susceptibility to enteric virus infection. Intestinal epithelial cells rhythmically express interferon-stimulated genes (ISGs) in anticipation of the feeding cycle, and disruption of this circadian programming — through clock gene deletion or jet lag — dramatically increases viral infection susceptibility in mouse models.
Why it matters: Chrono-immunology is an underappreciated dimension of host defense. The finding that antiviral immunity is gated by circadian rhythms has immediate implications for vaccine timing, antiviral drug scheduling, and understanding why some individuals are more susceptible to infections at certain times.
Why for Yiru: Circadian regulation of immune function connects to broader interests in how systemic and environmental factors shape immune cell states. This is relevant to understanding temporal dynamics of antitumor immunity and optimizing immunotherapy timing.
A Specialized CD107a+ Macrophage Subset Drives Selective Mycobacterial Phagocytosis
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2025.10.31.685620
macrophage phagocytosis mycobacterium CD107a innate immunity subset specialization
Summary: Identifies a specialized CD107a+ (LAMP1+) macrophage subset with hyperphagocytic capacity for Mycobacterium tuberculosis. This subset is distinguished by enhanced actin cytoskeleton regulators (Arp2/3) and NF-κB-driven pro-inflammatory signaling. CD107a surface expression strongly correlates with mycobacterial uptake in an actin-dependent, cytochalasin D-sensitive manner, revealing functional phagocytic specialization among macrophage populations.
Why it matters: Macrophage functional heterogeneity is poorly understood despite macrophages being central to infection, inflammation, and cancer. The discovery of a molecularly defined hyperphagocytic subset opens the door to therapeutic strategies that expand or activate this population for tuberculosis and potentially other intracellular pathogens.
Why for Yiru: Macrophage subset specialization is directly relevant to understanding tumor-associated macrophage (TAM) heterogeneity and function. The concept of molecularly defined phagocytic subsets could extend to how TAMs differentially clear tumor cells or debris in the TME.
An X-linked long non-coding RNA, PTCHD1-AS, and the core features of autism
Nature Published 2026-05-13 research article DOI: 10.1038/s41586-026-10515-6
lncRNA autism X-linked PTCHD1-AS neurodevelopment noncoding genome
Summary: Identifies PTCHD1-AS, an X-linked long non-coding RNA, as a key regulator of the core features of autism spectrum disorder. Using human iPSC-derived neurons and mouse models, the study shows that PTCHD1-AS regulates synaptic gene expression programs and that its loss recapitulates autism-relevant cellular and behavioral phenotypes. The antisense transcript functions in cis to regulate its neighboring protein-coding gene PTCHD1.
Why it matters: The noncoding genome's contribution to neurodevelopmental disorders remains largely unexplored. This study establishes a specific lncRNA as causal for autism core features, opening a new class of therapeutic targets and highlighting the importance of the noncoding genome in complex brain disorders.
Why for Yiru: Long non-coding RNA biology and its role in gene regulation is broadly relevant to understanding epigenetic and regulatory mechanisms in cancer and development. The cis-regulatory mechanism demonstrated here is a paradigm that likely extends to other disease contexts.
Cross-disciplinary watchlist
Other Fields
Pharmabiotics, Phocaeicola dorei, ameliorates cholestatic liver fibrosis by alleviating macrophage efferocytosis of neutrophils
Nature Communications Published 2026-05-17 research article DOI: 10.1038/s41467-026-73166-1
microbiome liver fibrosis macrophage efferocytosis pharmabiotics Phocaeicola dorei
Summary: Demonstrates that the gut commensal bacterium Phocaeicola dorei produces bioactive metabolites (pharmabiotics) that ameliorate cholestatic liver fibrosis. The mechanism involves modulation of macrophage efferocytosis of neutrophils in the liver — a clearance process whose dysregulation drives persistent inflammation and fibrosis. P. dorei supplementation reduces fibrosis markers in mouse models, suggesting a microbiome-targeted therapeutic strategy for chronic liver disease.
Why it matters: Liver fibrosis affects millions worldwide with limited therapeutic options. A microbiome-based approach using defined bacterial strains or their metabolites represents a novel therapeutic modality. The efferocytosis mechanism also connects gut-liver axis biology to fundamental macrophage biology.
Why for Yiru: Microbiome-immune interactions and macrophage biology are broadly relevant to understanding tissue homeostasis and inflammation. The gut-liver axis is a model for how distal microbial signals shape local immune function.
Structure-Led Exploration of the Metagenome Yields Novel RNA-Guided Nucleases with Broad PAM Diversity
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2026.03.27.714800
CRISPR metagenomics RNA-guided nucleases PAM diversity protein structure genome editing
Summary: Uses structure-guided mining of metagenomic datasets to discover novel RNA-guided nucleases with broad PAM diversity. Starting from structural predictions rather than sequence homology, the authors identify CRISPR-associated enzymes that recognize PAM sequences not covered by existing Cas9/Cas12 variants, substantially expanding the targeting range for genome editing applications.
Why it matters: PAM constraints remain a major limitation for CRISPR genome editing — many genomic sites are inaccessible to current enzymes. Structure-led discovery, rather than sequence-based homology searches, accesses a larger portion of the natural CRISPR diversity and could unlock editing at previously inaccessible sites.
Why for Yiru: CRISPR technology and genome editing tools are foundational to functional genomics. Expanded PAM targeting directly enables more comprehensive genetic screens and therapeutic editing approaches.
Bioactive Enhanced Adjuvant Chemokine Oligonucleotide Nanoparticles (BEACON) for Mucosal Vaccination Against Genital Herpes
bioRxiv Published 2026-05-15 preprint DOI: 10.1101/2025.07.31.667899
vaccine nanoparticle HSV-2 mucosal immunity CXCL9 CpG tissue-resident memory
Summary: Develops BEACON (Bioactive Enhanced Adjuvant Chemokine Oligonucleotide Nanoparticles), a self-assembling nanoparticle adjuvant formed by electrostatic complexation of CpG oligonucleotides with the chemokine CXCL9, for mucosal vaccination against genital herpes. Vaginal co-administration with HSV-2 glycoproteins following intramuscular priming enhances protection by increasing local tissue-resident memory T cells and luminal antibodies while reducing neutrophilic inflammation compared to standard CpG adjuvants, from the Iwasaki lab at Yale.
Why it matters: HSV-2 affects over 500 million people globally with no approved vaccine. A mucosal vaccine strategy that enhances tissue-resident memory — the first line of defense at viral entry sites — represents a rationally designed approach that addresses the unique challenges of genital mucosal immunity.
Why for Yiru: Vaccine design and mucosal immunology are relevant to understanding how tissue-specific immune responses can be therapeutically harnessed. The chemokine-nanoparticle approach has conceptual parallels to TME-targeted immunomodulation.
Sustaining microglial reparative function enhances stroke recovery
Nature Published 2026-05-13 research article DOI: 10.1038/s41586-026-10480-0
stroke microglia repair neuroinflammation brain injury TREM2
Summary: Identifies a mechanism to sustain microglial reparative function after stroke, enhancing recovery outcomes. Shows that after ischemic injury, microglia initially adopt a neuroprotective phenotype but progressively lose this reparative capacity. Sustaining this function — through modulation of key signaling pathways — improves neuronal survival, promotes tissue repair, and enhances functional recovery in mouse stroke models.
Why it matters: Stroke is a leading cause of disability with few treatments beyond acute thrombolysis. Sustaining endogenous microglial repair programs represents a paradigm shift from acute intervention to recovery-phase therapy. The concept of maintaining rather than inducing reparative cell states may apply broadly to tissue injury and degenerative diseases.
Why for Yiru: Microglial biology and the concept of sustaining reparative cell states connect to broader principles of immune cell plasticity in tissue microenvironments. Understanding how reparative programs are lost over time has implications for chronic inflammation and cancer.
Visualization of the complete preprimosome reveals the structural mechanisms governing DNA replication restart
Nature Communications Published 2026-05-16 research article DOI: 10.1038/s41467-026-73239-1
DNA replication preprimosome cryo-EM structural biology genome stability replication restart
Summary: Presents the first complete cryo-EM structure of the bacterial preprimosome — the multi-protein complex that restarts stalled DNA replication forks. The structure reveals how PriA, PriB, PriC, DnaT, and DnaB assemble into a functional replication restart machine, providing mechanistic insight into how cells rescue collapsed replication forks to maintain genome integrity.
Why it matters: Replication fork stalling and collapse are major sources of genome instability driving cancer and aging. Understanding the structural basis of replication restart provides a mechanistic framework for how cells prevent catastrophic fork failure and may reveal vulnerabilities in cancer cells with defective replication stress responses.
Why for Yiru: Genome stability mechanisms are fundamental to cancer biology. Replication stress and fork protection are particularly relevant to understanding how rapidly dividing cancer cells maintain genomic integrity under therapeutic pressure.