Research Radar — 2026-06-27
Methods & AI
Computational
OpenIO: An open framework for AI-native immunotherapy
Cancer Cell Published 2026-06-24 commentary DOI: 10.1016/j.ccell.2026.06.002
AI immunotherapy foundation model precision oncology generative AI
Summary: The authors propose Open Immune Oncology (OpenIO), an open framework integrating generative AI and multi-omics to advance precision oncology. By leveraging biological scaling laws and foundation models, the framework aims to transition immunotherapy from empirical screening to rational, AI-native engineering of therapeutic interventions.
Why it matters: Outlines a systematic path to systematize immunotherapy design through foundation models, potentially accelerating the discovery and optimization of immunotherapeutic interventions from target identification to clinical deployment.
Why for Yiru: Directly relevant to AI and foundation model interests in immunotherapy — the framework concept aligns with computational immunotherapy and digital twin approaches in precision oncology.
Deciphering protein mutation-phenotype linkages from CRISPR-based tiling mutagenesis screens
Cell Systems Published 2026-06-25 research article DOI: 10.1016/j.cels.2026.101651
CRISPR screen deep learning protein structure computational biology mutation-phenotype
Summary: He et al. present ProTiler-Mut, a computational framework that leverages CRISPR tiling mutagenesis screens to systematically decipher protein mutation-phenotype relationships. By integrating multi-condition phenotypes with 3D structural and interaction mapping, ProTiler-Mut reveals disease-associated hotspot substructures and rewired protein-protein interactions.
Why it matters: Integrates CRISPR perturbation screens with structural biology and deep learning to systematically map mutation-to-phenotype effects, providing a powerful new paradigm for functional genomics and variant interpretation.
Why for Yiru: Combines CRISPR screen data with computational modeling — relevant to functional genomics and variant effect prediction, with potential applications in cancer driver gene discovery.
High-throughput machine learning-aided antibody discovery for cell surface antigens
Cell Systems Published 2026-06-22 research article DOI: 10.1016/j.cels.2026.101645
machine learning antibody discovery deep learning drug discovery high-throughput screening
Summary: The authors present a minimalist synthetic Fab yeast display library encoding antigen recognition within a compact CDRH3-based module optimized for machine learning. Parallel screening across ten cell surface targets yields developable antibodies, while logistic regression rescues overlooked binders, establishing an ML-ready dataset and hybrid discovery framework that accelerates antibody generation.
Why it matters: Demonstrates a proof-of-concept for ML-integrated antibody discovery pipelines, showing how simple logistic regression can rescue binders missed by conventional screening and create ML-ready datasets for future optimization.
Why for Yiru: ML-driven antibody discovery relevant to computational drug design interests — the hybrid screening + ML framework could be extended to other therapeutic modalities.
scRADAR: Dissecting intratumoral drug response heterogeneity at single-cell resolution via mechanism-guided prototype routing
PLOS Computational Biology Published 2026-06-26 research article DOI: 10.1371/journal.pcbi.1014392
single-cell drug response deep learning cancer intratumoral heterogeneity
Summary: scRADAR is a mechanism-guided prototype routing framework for predicting drug-response phenotypes at single-cell resolution. It integrates metabolic and signaling pathway activities to form dual-view cellular representations, conditions predictions on drug mechanisms of action, and uses sparse prototype routing for interpretable predictions of drug resistance.
Why it matters: Bridges single-cell resolution with mechanism-aware deep learning to predict drug response, addressing a critical bottleneck in precision oncology: resolving which cell states within heterogeneous tumors will resist or respond to therapy.
Why for Yiru: Directly relevant — single-cell drug response prediction connects spatial transcriptomics, deep learning, and tumor microenvironment analysis, all core research interests.
From prediction to interpretation in computational pathology
Cancer Cell Published 2026-06-24 preview DOI: 10.1016/j.ccell.2026.06.003
AI computational pathology imaging deep learning biomarker spatial
Summary: This preview discusses PathPrism, a framework that demonstrates how interpretable spatial representations of tissue organization can support biomarker discovery, clinical prediction, and hypothesis generation from routine histopathology slides. It highlights the emerging transition from black-box prediction toward biologically meaningful interpretation in computational pathology.
Why it matters: Marks a shift toward interpretable AI in pathology, which is critical for clinical adoption and for enabling biological discovery from routine histology data.
Why for Yiru: Interpretable AI for spatial tissue analysis — relevant to computational pathology interests and spatial biomarker discovery from histology images.
Single-threshold–guided adaptive cancer therapy with partial-cycle treatment: A mechanistic and reinforcement learning analysis
PLOS Computational Biology Published 2026-06-26 research article DOI: 10.1371/journal.pcbi.1014457
cancer AI reinforcement learning computational modeling immunotherapy
Summary: Adaptive cancer therapy seeks to modulate aggressive treatment to preserve drug-sensitive tumor cells. This study proposes a single-threshold-guided adaptive therapy with partial surveillance-cycle treatment (AT-PSC), using mechanistic modeling and reinforcement learning to prolong time to progression by 402 days compared with conventional full-cycle adaptive therapy.
Why it matters: Provides a computational framework for optimizing adaptive therapy schedules, potentially improving cancer treatment outcomes while reducing toxicity from continuous high-dose therapy.
Why for Yiru: Reinforcement learning for treatment optimization — tangentially relevant to computational modeling interests in cancer.
Biomedical discoveries
Biomedicine
The transcription factor Eomes drives a stemness program in CD4+ T cells that promotes anti-tumor immunity in response to immunotherapy
Immunity Published 2026-06-23 research article DOI: 10.1016/j.immuni.2026.05.018
T cell immunotherapy tumor microenvironment stemness CD4+ T cells Eomes
Summary: The transcription factor Eomes orchestrates the differentiation and maintenance of an exhausted-like Th cell lineage that is transcriptionally and functionally distinct from conventional effector and memory Th subsets. This lineage is amplified by 4-1BB (CD137) stimulation, promoting effective Th-cell-mediated tumor control, and its progenitors exhibit self-renewal capacity that is conserved between mice and humans and selectively expanded by immune checkpoint inhibitors.
Why it matters: Identifies a novel CD4+ T cell lineage with stem-like properties that responds to checkpoint blockade and 4-1BB agonism, fundamentally expanding our understanding of CD4+ T cell differentiation and anti-tumor immunity beyond the conventional helper/effector paradigm.
Why for Yiru: Core interest area — T cell stemness programs, immunotherapy mechanisms, and CAR-T biology. The Eomes-driven CD4+ lineage has direct implications for designing T cell-based therapies with durable anti-tumor activity.
Molecular phenotypes and spatial archetypes: A new framework for cancer-associated fibroblasts
Cancer Cell Published 2026-06-24 review DOI: 10.1016/j.ccell.2026.06.001
tumor microenvironment CAF spatial cancer immunotherapy
Summary: Liu et al. propose a unifying framework that organizes cancer-associated fibroblasts into conserved molecular phenotypes and spatial archetypes. By linking cellular lineage, tissue context, and function, this model clarifies the long-standing heterogeneity in CAF biology and provides a foundation for developing precision strategies to selectively target pathogenic tumor-promoting stroma.
Why it matters: Provides a systematic, actionable framework for understanding CAF heterogeneity, which has been a major barrier to developing effective stroma-targeted cancer therapies.
Why for Yiru: Highly relevant — CAFs are a central component of the tumor microenvironment. The spatial archetype framework directly connects to spatial transcriptomics methodology and tumor microenvironment deconvolution interests.
Quantitative cytokine profiling of primary human macrophages reveals distinct single-cell modes of trained immunity
Cell Systems Published 2026-06-22 research article DOI: 10.1016/j.cels.2026.101648
macrophage trained immunity single-cell immune cytokine
Summary: Macrophages with prior training experiences mount stronger transcriptional responses to restimulation. O'Farrell et al. reveal gene-specific single-cell transcriptional changes that generate population-wide training phenotypes in human macrophages and highlight shifting time dynamics that underscore the transcriptional basis of trained immunity.
Why it matters: Provides mechanistic insight into trained immunity at single-cell resolution, revealing how prior exposure epigenetically programs macrophage responses with implications for vaccination, infection, and cancer immunotherapy.
Why for Yiru: Macrophage biology and single-cell analysis — directly relevant to tumor-associated macrophage characterization and understanding innate immune contributions to the tumor microenvironment.
Chronic stress unleashes an intratumor phage-fibroblast-B cell circuit to promote tumor growth
Cancer Cell Published 2026-06-24 research article DOI: 10.1016/j.ccell.2026.06.004
tumor microenvironment cancer immune CAF B cell microbiome
Summary: Bashir et al. show that chronic stress disrupts the gut barrier, enabling Enterococcus gallinarum translocation to tumors. There, Eg phage DNA activates cancer-associated fibroblasts via TLR9 to produce local glucocorticoids that suppress germinal-center B cell responses and accelerate tumor growth. Targeting this phage-CAF-B cell circuit restores anti-tumor immunity.
Why it matters: Reveals a novel gut-brain-tumor axis connecting chronic stress, microbiome translocation, phage DNA, CAF activation, and B cell suppression — uncovering an entirely unexpected mechanism of immune evasion.
Why for Yiru: Tumor microenvironment and CAF biology — the phage-CAF-B cell circuit is a novel and surprising mechanism connecting stress to immune suppression in tumors, with potential therapeutic interventions.
Monocytic niches escape T cell surveillance and promote Mycobacterium tuberculosis persistence in lymph nodes
Immunity Published 2026-06-22 research article DOI: 10.1016/j.immuni.2026.05.017
immune T cell macrophage infection immune evasion
Summary: Shamskhou et al. shed light on the paradox of lymph nodes as both sites of protective T cell responses and chronic M. tuberculosis infection. They show a shift from early dendritic cell-driven Th1 cell priming to monocyte-derived bacterial niches in the T cell zone that evade T cell recognition and promote persistent infection.
Why it matters: Explains how M. tuberculosis establishes chronic infection by creating specialized monocytic niches that physically and functionally escape T cell surveillance, revealing immune evasion principles relevant beyond infectious disease.
Why for Yiru: Immune evasion mechanisms and T cell-macrophage interactions — relevant to understanding how myeloid niches can shield pathogens (and potentially tumor cells) from T cell immunity.
Intratumoral B cells under stress
Cancer Cell Published 2026-06-24 commentary DOI: 10.1016/j.ccell.2026.05.015
tumor microenvironment B cell cancer immune
Summary: This commentary discusses findings showing that stress-induced corticosterone suppresses germinal center B cell responses and impairs anti-tumor immunity. It identifies gut microbiota-derived phage DNA as a key trigger of TLR9 activation and corticosterone secretion by cancer-associated fibroblasts, revealing an intricate bacteria-driven immune-suppression mechanism.
Why it matters: Highlights the emerging and underappreciated role of intratumoral B cells in cancer immunity and how neuroendocrine stress signals can suppress this axis through CAF-mediated mechanisms.
Why for Yiru: B cell biology in the tumor microenvironment — complements the chronic stress article and connects to broader TME and spatial biology interests.
Cross-disciplinary watchlist
Other Fields
Mapping the spatial landscape of extracellular vesicles in tissues with Spatial-EV-seq
Nature Biotechnology Published 2026-06-25 research article DOI: 10.1038/s41587-026-03206-0
spatial transcriptomics extracellular vesicles tumor microenvironment method
Summary: Wen, Na, Lu, Zhang, Yang et al. present Spatial-EV-seq, a method for spatially resolved profiling of extracellular vesicles (EVs) in tissues while preserving their native distribution. By integrating EV mapping with spatial transcriptomics, the method enables location-specific analysis of EVs and their communication networks within the tissue microenvironment.
Why it matters: Adds an entirely new modality to the spatial biology toolkit — extracellular vesicle mapping — enabling researchers to study how cells communicate through EV cargo in their native tissue context.
Why for Yiru: Directly relevant to spatial transcriptomics and tumor microenvironment interests. This method opens up the possibility of studying intercellular communication networks spatially, which could reveal new mechanisms of tumor-immune cell crosstalk.
Near-perfect genome sequencing in medical genetics
Nature Genetics Published 2026-06-25 perspective DOI: 10.1038/s41588-026-02645-4
genomics AI sequencing precision medicine
Summary: This Perspective introduces near-perfect genome sequencing, which encompasses diploid genome assembly, pangenome references, and AI-driven variant interpretation, and proposes a roadmap toward its clinical implementation in medical genetics.
Why it matters: Lays out a vision for how advances in sequencing technology, pangenomics, and AI could make comprehensive and accurate genome interpretation a clinical reality.
Why for Yiru: Genomics and AI in precision medicine — relevant for understanding how sequencing technology and variant interpretation are evolving toward clinical deployment.
Three decades of cancer genetics
Nature Genetics Published 2026-06-25 perspective DOI: 10.1038/s41588-026-02657-0
cancer genetics genomics historical perspective
Summary: In this Q&A, Andrew Futreal reflects on three decades of cancer genetics, from the initial cloning of BRCA1 in 1994 to the outstanding questions in understanding how molecular variation drives clinical phenotypes in cancer patients.
Why it matters: Offers a retrospective on the evolution of cancer genetics from a pioneer in the field, highlighting both progress and the challenges that remain in translating genomic discovery to patient benefit.
Why for Yiru: Historical perspective on cancer genetics — light reading that contextualizes the field's trajectory.
Evaluating the robustness and readiness of large frontier models in health AI applications
Nature Medicine Published 2026-06-25 research article DOI: 10.1038/s41591-026-04501-8
AI foundation model clinical AI robustness evaluation
Summary: This adversarial evaluation of leading AI models uncovers significant gaps between benchmark success and real-world robustness, highlighting fundamental limitations in current health AI benchmarks and their ability to capture clinically relevant performance. The findings raise important questions about the readiness of large frontier models for clinical deployment.
Why it matters: Calls attention to the critical gap between impressive benchmark performance and actual clinical robustness of large AI models, with implications for regulatory approval and patient safety.
Why for Yiru: Essential reading for anyone building AI tools for biomedical applications — the robustness gap is a critical concern for translating foundation models into clinical and research use.