Research Radar — 2026-05-02
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 graphs transcriptomics perturbation prediction deep learning
Summary: TxPert leverages multiple knowledge graphs to predict single transcriptomic perturbation effects, achieving accuracy approaching split-half experimental reproducibility for unseen perturbations.
Why it matters: Predicting transcriptional responses to perturbations is a core challenge in drug discovery and functional genomics. Knowledge-graph approaches that generalize to unseen conditions could dramatically reduce the experimental burden of perturb-seq screens.
Why for Yiru: Directly relevant to computational methods for predicting biological perturbation outcomes — a capability essential for digital twin AI and in silico treatment modeling.
Single-cell foundation models reveal context-sensitive cancer programmes under subtype shift
bioRxiv (bioinformatics) Published 2026-04-28 preprint DOI: 10.64898/2026.04.28.721114
single-cell foundation models cancer biology transfer learning representation learning
Summary: Single-cell foundation models are applied to uncover cancer gene programmes that shift in context-dependent ways across tumor subtypes, demonstrating how pretrained representations can reveal biologically meaningful variation.
Why it matters: Foundation models pretrained on massive single-cell atlases are becoming a new paradigm for analyzing any scRNA-seq dataset. Understanding how their representations behave across cancer subtypes is critical for clinical translation.
Why for Yiru: Core interest area: single-cell foundation models, representation learning, and cancer biology. This paper bridges computational methodology with translational cancer insights.
CHAMPOLLION: Robust Multi-Omics Integration via Inverse Optimal Transport Using Paired Cells
bioRxiv (bioinformatics) Published 2026-04-28 preprint DOI: 10.64898/2026.04.28.721317
multi-omics integration optimal transport single-cell computational methods
Summary: CHAMPOLLION uses inverse optimal transport to robustly integrate multiple omics modalities from paired single-cell measurements, addressing the challenge of aligning disparate data types.
Why it matters: Multi-omics integration at single-cell resolution is a bottleneck in spatial and single-cell biology. Optimal transport provides a principled mathematical framework that can handle the heterogeneity and noise of real data.
Why for Yiru: Multi-omics integration and optimal transport are directly aligned with computational immunology and tumor microenvironment analysis — key tools for Boss's research.
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 chromatin accessibility lymphoma FFPE tissue histone modifications
Summary: epi-Patho-DBiT enables spatial mapping of chromatin accessibility and histone modifications in archived FFPE human lymphoma tissues, revealing epigenetic drivers of lymphoma development, progression, and transformation.
Why it matters: Spatial profiling of the epigenome in clinical FFPE samples opens the door to retrospective studies linking tissue architecture to gene regulation in cancer. This is a major advance for translational spatial biology.
Why for Yiru: Spatial omics technology development, particularly methods applicable to clinical archived samples, is directly relevant to tumor microenvironment research and computational pathology.
Reconstructing True 3D Spatial Omics at Single-Cell Resolution
bioRxiv (bioinformatics) Published 2026-04-28 preprint DOI: 10.64898/2026.04.28.721395
spatial omics 3D reconstruction single-cell resolution computational methods
Summary: A computational framework for reconstructing true three-dimensional spatial omics maps at single-cell resolution from serial tissue sections.
Why it matters: Most spatial transcriptomics is inherently 2D, but tissues are 3D. True 3D reconstruction at single-cell resolution would transform our understanding of tissue architecture and cell-cell communication.
Why for Yiru: Spatial omics and 3D tissue reconstruction are essential for modeling the tumor microenvironment — directly applicable to Boss's computational immunology work.
Biomedical discoveries
Biomedicine
Distinct in vivo dynamics of donor-derived stem cell memory CAR T cells post-allogeneic HSCT relapse
Cell Published 2026-04-30 research article DOI:
CAR T cell therapy stem cell memory allogeneic HSCT immunotherapy T cell biology
Summary: Donor-derived CD8+ CAR TSCM cells exhibit enhanced expansion and a favorable safety profile, inducing complete responses at low doses without lymphodepletion. Their distinctive in vivo behavior and differentiation trajectory establish TSCM cells as a robust and safe platform for next-generation CAR T cell therapy.
Why it matters: This study provides direct clinical evidence that stem cell memory CAR T cells can achieve potent anti-tumor responses with reduced toxicity, addressing two major limitations of current CAR T therapies.
Why for Yiru: CAR T cell biology and immunotherapy are core interests. The TSCM phenotype and in vivo dynamics directly inform Boss's computational modeling of immune cell states in the tumor microenvironment.
Activated T cell extracellular vesicle DNA transfer enhances antigen presentation and anti-tumor immunity
Cancer Cell Published 2026-04-30 research article DOI:
T cell extracellular vesicles antigen presentation immunotherapy tumor immunity DNA transfer
Summary: Activated T cell-derived extracellular vesicles (ATEVs) transfer genomic DNA enriched in antigen processing and presentation genes into dendritic and tumor cells via granzyme B-mediated nuclear delivery. This mechanism enhances anti-tumor immunity and synergizes with PD-1 blockade to overcome immune evasion.
Why it matters: This uncovers a previously unappreciated intercellular communication mechanism — T cells don't just kill; they also transfer functional DNA that boosts antigen presentation in neighboring cells. ATEVs represent a novel acellular immunotherapy platform.
Why for Yiru: T cell biology, antigen presentation, and tumor-immune interactions are all core areas. The EV-mediated DNA transfer mechanism could inspire computational models of intercellular communication in the TME.
Tissue tension fosters macrophage-driven lipid peroxidation-induced DNA damage
Cancer Cell Published 2026-04-30 research article DOI:
tumor microenvironment macrophage biology DNA damage lipid peroxidation tissue mechanics
Summary: Stromal fibrosis drives epithelial STAT3-mediated chemokine secretion to recruit macrophages. Under elevated tissue tension, macrophages undergo lipid peroxidation, generating aldehydes that induce epithelial DNA damage and promote tumor progression.
Why it matters: This study mechanistically connects three pillars of cancer — fibrosis, inflammation, and DNA damage — through tissue mechanics. It identifies a biophysical pathway by which the TME directly drives mutagenesis.
Why for Yiru: Macrophage biology in the tumor microenvironment, tissue mechanics, and the link between inflammation and DNA damage are all highly relevant to Boss's computational immunology and cancer biology interests.
A conserved re-epithelialization program underlies malignancy in pancreatic ductal adenocarcinoma
Cancer Cell Published 2026-04-30 research article DOI:
pancreatic cancer tumor-stroma crosstalk FOSL1 CAFs wound healing
Summary: A conserved cutaneous wound-healing programme (MP10), driven by FOSL1 and distinct from EMT, underlies PDAC malignancy. CTHRC1-high myCAFs promote FOSL1 expression through EGFR signaling, revealing targetable tumor-stroma crosstalk.
Why it matters: This redefines PDAC progression as co-opting a wound repair program rather than a developmental EMT. The EGFR-mediated fibroblast-tumor crosstalk is a druggable vulnerability.
Why for Yiru: Tumor-stroma interactions, pancreatic cancer biology, and the transcriptional programmes driving malignancy are all relevant to computational modeling of the TME.
Genetic variation reveals a homeotic long noncoding RNA that modulates human hematopoietic stem cells
Cell Published 2026-05-01 research article DOI:
lncRNA hematopoietic stem cells HOXA9 blood cancer genetic variation
Summary: A blood cancer-protective inherited variant in the homeotic lncRNA HOTSCRAMBL alters HOXA9 splicing, fine-tuning HSC self-renewal while constraining HOXA-driven blood cancers.
Why it matters: This identifies a human genetic variant that naturally balances HSC fitness against cancer risk through lncRNA-mediated splicing regulation — a beautiful example of human genetics illuminating fundamental stem cell biology.
Why for Yiru: Hematopoietic stem cell biology and the genetic basis of blood cancers are relevant to Boss's broader interests in computational immunology and cancer genomics.
Cross-disciplinary watchlist
Other Fields
Call your AI agent
Nature Methods Published 2026-05-01 editorial DOI: 10.1038/s41592-026-03088-9
AI agents large language models scientific automation AI in research
Summary: Nature Methods editorial discusses the emergence of AI agent systems built on large language models that can conduct autonomous scientific analyses, urging a balance of curiosity, excitement, and skepticism when choosing among offerings.
Why it matters: AI agents are rapidly moving from demo to deployment in scientific workflows. Understanding their capabilities, limitations, and risks is essential for any computational biologist.
Why for Yiru: AI agents for scientific research directly intersect with Boss's interests in biomedical AI and computational methods. The editorial frames the landscape of tools that could augment or automate parts of the research pipeline.
Skill-Augmented Frontier Agents Nearly Saturate BixBench-Verified-50
bioRxiv (bioinformatics) Published 2026-04-28 preprint DOI: 10.64898/2026.04.28.721523
AI agents benchmarking frontier models skill augmentation
Summary: Skill-augmented frontier AI agents achieve near-saturation performance on the BixBench-Verified-50 benchmark, demonstrating rapid progress in agentic AI capabilities.
Why it matters: Benchmark saturation by AI agents signals that agentic AI is maturing rapidly. Understanding what benchmarks measure and what they miss is critical for evaluating AI tools in scientific contexts.
Why for Yiru: AI agent benchmarking is relevant to evaluating which AI systems might be useful for biomedical research automation, and to understanding the frontier of general AI capabilities.