Research Radar — 2026-07-23
Methods & AI
Computational
Expanding the scope of protein language modeling to protein-protein interactions with MSA Pairformer
Cell Published 2026-07-21 AI_model DOI: 10.1016/j.cell.2026.06.029
deep learning protein language model protein-protein interactions computational biology AI
Summary: MSA Pairformer is a parameter-efficient protein language model that generalizes to protein-protein interactions despite being trained exclusively on monomer sequences. By extending the Pairformer architecture to handle co-evolutionary signals across interacting protein families, it enables accurate prediction of interface variant effects and subfamily-specific homo-oligomeric binding modes, bridging a critical gap between single-chain protein modeling and the reality that most proteins function through interactions.
Why it matters: Bridges the gap between single-chain protein language models and complex prediction, enabling AI-driven design of protein complexes with therapeutic and biotechnological applications without requiring training data on protein interfaces.
Why for Yiru: The extension to pairwise interactions demonstrates a principled approach that could inspire similar advances in modeling ligand-receptor interactions in spatial transcriptomics and cell-cell communication networks in the tumor microenvironment.
Precise DNA base editing using AlphaFold3-based contact modelling
Nature Published 2026-07-22 AI_model DOI: 10.1038/s41586-026-10794-z
base editing AlphaFold deep learning CRISPR protein structure
Summary: ContactSeek is an AlphaFold3-driven model that improves the precision of genome-editing tools by using contact maps to predict off-target effects of base editors with high accuracy. The approach demonstrates how protein structure prediction can directly enhance the safety and specificity of therapeutic genome editing.
Why it matters: Demonstrates that AlphaFold3-based contact modelling can directly improve the safety profile of base editors, addressing a key clinical barrier for therapeutic genome editing by enabling more precise prediction of off-target effects.
Why for Yiru: Highly relevant to my deep learning and genome editing interests — the ContactSeek framework represents a template for how AI-driven protein structure prediction can improve molecular tool design, a paradigm applicable to engineering improved CAR constructs and other therapeutic proteins.
AI-redesigned starting points and outcomes enhance protein evolution
Nature Published 2026-07-22 AI_model DOI: 10.1038/s41586-026-10820-0
protein engineering AI directed evolution computational design deep learning
Summary: A workflow using artificial intelligence-redesigned starting points to evolve enzymes with improved properties compared with those evolved from natural proteins. The study establishes that AI-generated protein scaffolds can serve as superior starting templates for directed evolution, yielding enzymes with enhanced catalytic efficiency and stability.
Why it matters: Establishes a paradigm where AI-designed protein scaffolds outperform natural proteins as starting points for directed evolution, significantly accelerating the development of enzymes with industrial and therapeutic applications.
Why for Yiru: Relevant to my AI and protein engineering interests — the concept of AI-optimized starting points for evolution could extend to designing improved immune receptors, CAR constructs, or other therapeutic proteins with enhanced functionality.
IOBRpy enables agentic multi-omics decoding of anti-tumor immunity
bioRxiv Published 2026-07-22 computational_tool DOI: 10.64898/2026.07.17.739055v1
multi-omics tumor immunity bioinformatics AI agent immunotherapy
Summary: IOBRpy is a Python toolkit driven by an innovative AI dual-agent layer for automated, standardized immuno-oncology workflows. It integrates upstream quality control, transcript quantification, and downstream TME parsing including signature scoring, ligand-receptor crosstalk, cellular deconvolution, HLA typing, and TCR/BCR repertoire reconstruction. Deployed across IMvigor210 and OAKPOPLAR cohorts, it captures multi-dimensional prognostic insights and treatment-stratified allele-specific survival associations.
Why it matters: Provides a unified, AI-agent-driven platform for comprehensive immuno-oncology analysis from raw sequencing data, addressing the fragmentation of existing bioinformatics tools and enabling reproducible multi-omics immune profiling at scale.
Why for Yiru: Directly relevant to my multi-omics and tumor immunity work — the integrated HLA typing, TCR repertoire reconstruction, and TME deconvolution within a single framework would streamline my analyses of immune microenvironment interactions in spatial and single-cell datasets.
scLEMBAS: Context-Aware Modeling of Signaling Pathway Activity at Single-Cell Resolution
bioRxiv Published 2026-07-22 AI_model DOI: 10.64898/2026.07.20.739670v1
single-cell signaling pathway neural network perturbation computational biology
Summary: scLEMBAS is a context-aware, gray-box neural network that models signaling pathway activity at single-cell resolution while preserving mechanistic grounding. It encodes a prior-knowledge network of protein-protein interactions as a recurrent neural network with learnable edge weights corresponding to signaling strengths. An adversarial approach enables single-cell counterfactual prediction of what a given cell's transcription factor activity would be under different perturbations or contexts.
Why it matters: Bridges predictive single-cell modeling with mechanistic interpretability, enabling counterfactual reasoning about perturbation responses at single-cell resolution — a capability essential for understanding how context shapes cellular responses in disease and therapy.
Why for Yiru: Highly relevant to my single-cell and perturbation biology interests — the ability to predict context-dependent perturbation responses at single-cell resolution is directly applicable to modeling how tumor microenvironment signals modulate immune cell function and therapy response.
SPgen: Proteome-wide Spatial Proteomics generation using multi-modality foundation models
bioRxiv Published 2026-07-21 AI_model DOI: 10.64898/2026.07.16.739037v1
spatial proteomics foundation model multi-modal AI protein prediction
Summary: SPgen is a multi-modal foundation-model framework for proteome-wide spatial protein prediction that integrates protein sequences, functional annotations, transcriptomic profiles, and spatial information to learn transferable representations. It accurately reconstructs measured spatial patterns, reduces measurement noise, and enables proteome-wide spatial prediction beyond experimentally profiled proteins, addressing the limited protein coverage of current spatial proteomics technologies.
Why it matters: Overcomes a fundamental limitation of spatial proteomics — the inability to profile the full proteome — by using multi-modal foundation models to predict protein abundance from transcriptomic and spatial data, dramatically expanding the molecular scope of spatial analyses.
Why for Yiru: Directly relevant to my spatial biology and foundation model interests — the multi-modal spatial prediction approach could be adapted to predict protein-level immune markers in my spatial transcriptomics datasets, bridging the gap between RNA and protein measurements in the tumor microenvironment.
Biomedical discoveries
Biomedicine
Single-nucleus multimodal spatial transcriptomics reveals spatial colocalization of neoantigen-expressing tumor cells and cognate T cells
Nature Biotechnology Published 2026-07-22 spatial_omics DOI: 10.1038/s41587-026-03194-1
spatial transcriptomics neoantigen T cell tumor microenvironment single-cell multimodal
Summary: A single-nucleus multimodal spatial transcriptomics framework enables mapping of neoantigen-expressing tumor cells and tumor-infiltrating T cells at single-cell resolution, combined with T cell receptor sequencing. The approach reveals distinct immune niches in the tumor microenvironment and their properties that determine tumor-specific responses, providing direct evidence for spatial colocalization of neoantigen-expressing tumor cells with cognate T cells.
Why it matters: Provides the first direct spatial evidence of neoantigen-T cell colocalization in human tumors at single-cell resolution, establishing a framework for understanding how spatial organization of the immune microenvironment governs anti-tumor immunity.
Why for Yiru: Directly at the intersection of my core research interests in spatial transcriptomics, neoantigens, and the tumor microenvironment — the multimodal spatial approach combining transcriptomics with TCR sequencing is exactly the type of integrative analysis that could reveal how immune niches shape immunotherapy responses.
Tertiary lymphoid structures harbour stem-like tumour-specific T cells
Nature Published 2026-07-22 cancer_immunology DOI: 10.1038/s41586-026-10808-w
tertiary lymphoid structures T cell tumor microenvironment renal cell carcinoma immunotherapy
Summary: Renal cell carcinoma tumours containing tertiary lymphoid structures (TLSs) are enriched for exhausted CD8+ T cells, including tumour-specific clonotypes with stem-like progenitor features and reduced terminal exhaustion. This study reveals that TLSs serve as niches for maintaining stem-like tumour-specific T cells with progenitor potential, providing mechanistic insight into why TLSs correlate with favourable immunotherapy outcomes.
Why it matters: Provides mechanistic understanding of why tertiary lymphoid structures predict immunotherapy response by showing they harbour stem-like tumour-specific T cells, establishing TLSs as functional immune hubs rather than merely structural markers of inflammation.
Why for Yiru: Directly relevant to my tumor microenvironment and immunotherapy interests — the stem-like T cell population in TLSs represents a potential cellular target for combination strategies that preserve progenitor T cell function while enhancing anti-tumor immunity.
Quiescent tumor cells shape the immunosuppressive microenvironment in pancreatic cancer
Nature Communications Published 2026-07-21 cancer_immunology DOI: 10.1038/s41467-026-75883-z
pancreatic cancer tumor microenvironment CAR-T quiescent cells Epiregulin macrophage immunotherapy resistance
Summary: A rare population of quiescent pancreatic cancer cells increases after CAR-T therapy, expressing high levels of Epiregulin (EREG) which induces an immunosuppressive tumor microenvironment via ErbB4-expressing macrophages. Targeting EREG enhances CAR-T sensitivity, identifying a resistance mechanism driven by therapy-induced quiescence and macrophage reprogramming in one of the most immunotherapy-resistant solid tumors.
Why it matters: Reveals a therapy-induced resistance mechanism in which quiescent tumor cells actively remodel the microenvironment through EREG-ErbB4 macrophage signaling, providing a targetable axis to improve CAR-T efficacy in pancreatic cancer.
Why for Yiru: Highly relevant to my CAR-T, macrophage, and tumor microenvironment interests — the finding that quiescent tumor cells communicate with macrophages to drive immunosuppression identifies a targetable vulnerability that could be exploited in combination with CAR-T therapy for solid tumors.
Blocking the m6Am methyltransferase PCIF1 releases STAT1-mediated Th1 immunity to potentiate cancer immunotherapy
Nature Communications Published 2026-07-22 cancer_immunology DOI: 10.1038/s41467-026-75269-1
immunotherapy epigenetics T cell m6A STAT1 cancer
Summary: PCIF1 is the sole enzyme catalyzing the N6-methyladenosine (m6A) modification on mRNA. Genomic deletion of PCIF1 specifically in T cells releases STAT1 translation, promoting Th1 immunity and potentiating cancer immunotherapy. The study identifies PCIF1 as an T cell-intrinsic epigenetic checkpoint that restricts anti-tumor immunity through post-transcriptional regulation of STAT1.
Why it matters: Identifies a previously unrecognized T cell-intrinsic epigenetic checkpoint that can be targeted to enhance anti-tumor immunity, opening new avenues for combination immunotherapy strategies based on RNA modification.
Why for Yiru: Relevant to my immunotherapy and T cell biology interests — the discovery that RNA methylation acts as a T cell checkpoint reveals a new layer of immune regulation that could be leveraged to enhance CAR-T and checkpoint immunotherapy efficacy.
Localized PD-1 CAR T therapy reprograms neuroinflammation
Cell Published 2026-07-21 immuno_therapy DOI: 10.1016/j.cell.2026.06.036
CAR-T immunotherapy neuroinflammation PD-1 T cell
Summary: PD-1+ T follicular helper-like cells drive B cell-associated pathology in multiple sclerosis. Programmable PD-1-targeting CAR T cells selectively eliminate these pathogenic CD4 T cells while delivering IL-10 at sites of inflammation, thereby suppressing neuroinflammation across preclinical models. The study expands the therapeutic scope of CAR T cells beyond oncology to autoimmune and inflammatory conditions.
Why it matters: Expands the therapeutic scope of CAR T cells beyond cancer to neuroinflammatory disease, demonstrating that localized delivery can effectively reprogram tissue microenvironments without the systemic toxicity of traditional immunosuppressive therapies.
Why for Yiru: Highly relevant to my CAR-T and immunotherapy interests — the concept of localized CAR T delivery to reprogram tissue microenvironments could be adapted for solid tumors, where regional administration may overcome trafficking barriers and improve therapeutic index in the tumor microenvironment.
Mechanisms regulating combination effect of antibody-drug conjugates and cancer immunotherapy
bioRxiv Published 2026-07-22 cancer_immunology DOI: 10.64898/2026.07.17.738956v1
ADC immunotherapy T cell cancer combination therapy antibody
Summary: Antibody-drug conjugates (ADCs) represent a transformative class of cancer therapeutics, yet the mechanisms underlying their synergy with immunotherapy remain poorly understood. This study demonstrates that ADC/TCE combinations produce synergistic antitumor activity via a TNF/TNFR axis, while ADC/CPI combinations operate through an M6PR-dependent granzyme B pathway. ADC-induced autophagy is identified as a unifying, target- and payload-agnostic mechanism that sensitizes tumor cells to T cell-mediated killing.
Why it matters: Provides a mechanistic framework for rationally designing ADC and immunotherapy combination strategies, identifying distinct pathways for T cell engager versus checkpoint inhibitor combinations that could guide clinical trial design.
Why for Yiru: Directly relevant to my immunotherapy and combination therapy interests — understanding the distinct mechanisms of ADC synergy with T cell engagers versus checkpoint inhibitors informs rational combination design for my own translational research.
Cross-disciplinary watchlist
Other Fields
CRISPR–Cas regulates expression of embedded anti-phage defence systems
Nature Published 2026-07-22 microbiology DOI: 10.1038/s41586-026-10833-9
CRISPR microbiology gene regulation phage defence
Summary: CRISPR–Cas systems transcriptionally tune diverse innate defences using CRISPR RNA-like guides to balance antiviral protection with fitness. When compromised, these systems hyperactivate the embedded defences, establishing a layered bacterial 'immunity guard' network. This work reveals that CRISPR systems serve dual roles as adaptive immune systems and as regulators of broad-spectrum antiviral defence.
Why it matters: Reveals a previously unrecognized regulatory function of CRISPR-Cas systems beyond adaptive immunity, showing they act as master regulators of a layered bacterial antiviral defence network.
Why for Yiru: While primarily in microbiology, the finding that CRISPR systems have regulatory functions beyond DNA targeting is broadly relevant to my CRISPR interests and expands the conceptual framework for how these systems could be engineered for new applications.
Subnuclear genome compartmentalization controls bivalent chromatin activity
Nature Published 2026-07-22 epigenomics DOI: 10.1038/s41586-026-10832-w
chromatin epigenomics genome organization gene regulation
Summary: The spatial location of a gene within the nucleus is essential for understanding its epigenomic regulation. This study reveals how subnuclear genome compartmentalization controls bivalent chromatin activity — genes poised for activation or repression — demonstrating that nuclear positioning is a functional determinant of gene expression potential.
Why it matters: Establishes a direct functional link between the spatial position of genes within the nucleus and their epigenetic state, advancing our understanding of how 3D genome organization regulates gene expression.
Why for Yiru: Relevant to my genomics and chromatin interests — the principles of nuclear compartmentalization in regulating chromatin states could inform how spatial organization of the genome influences immune cell identity and function in the tumor microenvironment.
Genetic background sets the trajectory of experimental cancer evolution
Nature Published 2026-07-22 cancer_genetics DOI: 10.1038/s41586-026-10821-z
cancer evolution genetics mouse models tumor biology
Summary: Experimentally replaying tumour evolution in divergent mouse strains reveals the importance of interactions between genetic ancestry and acquired cancer-driving mutations in shaping the earliest stages of cancer development. The study demonstrates that genetic background profoundly influences which mutations drive tumour progression and the evolutionary trajectory of cancer.
Why it matters: Demonstrates that host genetic background is a major determinant of cancer evolution, with implications for understanding population-level differences in cancer susceptibility and progression.
Why for Yiru: Relevant to my cancer biology interests — the finding that genetic background shapes mutation trajectories highlights the importance of incorporating host genetics when studying tumor-immune interactions and immunotherapy responses.
Efficient and precise programmable DNA knock-in without double-strand breaks
Nature Published 2026-07-22 gene_editing DOI: 10.1038/s41586-026-10819-7
gene editing knock-in CRISPR DNA repair biotechnology
Summary: Kilobase-scale nickase-targeting (KNIT) editing enables single-nick- based DNA insertion across genomic loci and cell types, supporting insertion of DNA fragments exceeding 10 kb with high efficiency and minimal unwanted off-target effects. The approach avoids the risks associated with double-strand break repair, representing a major advance in precision genome engineering.
Why it matters: Overcomes a fundamental limitation of current gene editing technologies by enabling large DNA insertions without double-strand breaks, dramatically improving safety and opening new possibilities for gene therapy and synthetic biology.
Why for Yiru: Highly relevant to my gene editing interests — the KNIT approach for scarless, efficient large DNA insertion could enable more sophisticated engineering of CAR constructs and synthetic gene circuits for immunotherapy applications.
Prior therapy defines mutation profiles in childhood cancer at relapse
Nature Published 2026-07-22 cancer_genomics DOI: 10.1038/s41586-026-10803-1
childhood cancer mutation chemotherapy tumor evolution genomics
Summary: Chemotherapy, particularly with platinum-based drugs, is associated with substantial, rapidly detectable mutagenesis in childhood cancers, tripling private mutation signatures and doubling mutation burden. This large-scale genomic study reveals therapy-induced tumour evolution and identifies opportunities for safer, de-escalated treatment strategies.
Why it matters: Provides direct genomic evidence that chemotherapy drives mutagenesis in childhood cancers, with implications for treatment de-escalation and the development of therapies that minimize secondary mutations.
Why for Yiru: Relevant to my cancer genomics interests — the finding that prior therapy profoundly shapes the mutational landscape at relapse highlights the importance of considering treatment history when analyzing tumor-immune interactions and predicting immunotherapy responses.
The genomic position of an enhancer modulates bursting dynamics of the cognate promoter
Nature Genetics Published 2026-07-21 research_briefing DOI: 10.1038/s41588-026-02693-w
enhancer promoter gene regulation transcription genomics
Summary: In a genomic locus with minimal complexity, the distance between promoter and enhancer modulates the frequency of clustered transcriptional bursts. Beyond nucleotide sequence, the relative genomic position of promoter and enhancer controls mRNA production and cell-to-cell transcriptional variability, contributing to the precision of transcriptional responses.
Why it matters: Reveals that genomic distance itself — not just sequence — is a regulatory parameter controlling transcriptional bursting and variability, adding a new dimension to our understanding of enhancer-promoter communication.
Why for Yiru: The principle that spatial positioning within the genome controls gene expression noise is conceptually relevant to understanding how chromatin architecture in the tumor microenvironment may influence immune gene expression and cellular heterogeneity.