Research Radar — 2026-08-26
Methods & AI
Computational
Deep learning aging marker from retinal images unveils sex-specific clinical and genetic signatures
Nature Communications Published 2026-08-26 Research article DOI: 10.1038/s41467-026-77102-1
biomedical AI digital phenotyping aging retinal imaging genetics
Summary: The authors develop a retinal age predictor and show that accelerated retinal aging is associated with age-related diseases, mortality, and distinct sex-specific genetic and biological signatures.
Why it matters: The work turns an image-derived phenotype into a measurable biological-age signal that can be tested against clinical outcomes and genetic mechanisms, rather than treating retinal imaging as a benchmark alone.
Why for Yiru: It is a strong biomedical-AI and digital-phenotyping example for connecting image models to outcomes and mechanisms, with ideas transferable to multimodal disease modeling.
ClustoCell reveals cell states and their markers from single-cell transcriptomes
bioRxiv (Bioinformatics) Published 2026-08-25 Preprint DOI:
spatial omics single-cell analysis computational immunology cell states cell annotation
Summary: ClustoCell identifies cell states from within-cell transcriptional architecture by stratifying gene expression into high and medium tiers and constructing similarity graphs. Across 450 datasets and more than 24 million cells, it reports high concordance with expert annotations, stable and coherent states, rare and transitional populations, and an immunotherapy application linking pretreatment immune circuits to PD-1 responsiveness.
Why it matters: Reference-free, interpretable state identification could reduce sensitivity to atlas choice and preprocessing while making rare or transitional populations easier to resolve, although the broad concordance claim needs careful benchmark inspection.
Why for Yiru: The method is directly relevant to spatial and single-cell workflows and computational immunology, especially when reference atlases are incomplete or biological states are continuous.
MultiFlow: coupled flow matching for predicting single-cell multiomic perturbation responses in unseen cellular contexts
bioRxiv (Bioinformatics) Published 2026-08-25 Preprint DOI:
spatial omics multiomics perturbation biology RNA-ATAC flow matching
Summary: MultiFlow uses coupled flow matching to jointly generate and predict paired gene-expression and chromatin-accessibility responses, conditioned on perturbation and control-derived cellular state. The study reports coordinated RNA-ATAC generation and perturbation prediction in unseen cellular contexts, including preserved peak-gene effects and cross-modal neighborhood structure.
Why it matters: Joint modeling tests whether perturbation models preserve cross-modal coordination rather than independently reproducing each modality, a key requirement for credible multiomic response inference.
Why for Yiru: This is a direct methodological fit for spatial multiomics and perturbation modeling, with a useful design principle for transferring response predictions across cellular contexts.
Benchmark pitfalls expose need for expert-guided spatial clustering
Nature Methods Published 2026-08-24 Research perspective DOI: 10.1038/s41592-026-03193-9
spatial omics benchmarking reproducibility spatial clustering histology
Summary: The article argues that benchmarking spatially resolved transcriptomics tools is limited by reproducibility, data availability, and evaluation strategies that fail to represent multiscale spatial biology. It reports that complementing computational consensus strategies with histology and cell-biology expert feedback can overcome these limitations and accelerate biological discovery.
Why it matters: It frames expert-guided evaluation as a practical guard against benchmarks that reward consensus without biological validity, providing a standard for judging spatial methods rather than another clustering algorithm to adopt uncritically.
Why for Yiru: This is highly relevant to method development and evaluation in spatial omics, where reproducibility and multiscale biological validity should be explicit acceptance criteria.
Biomedical discoveries
Biomedicine
Dissecting context-dependent cancer vulnerabilities using Perturb-seq
bioRxiv (Cancer Biology) Published 2026-08-25 Preprint DOI:
computational immunology single-cell perturbation oncology functional genomics cancer vulnerabilities
Summary: The authors build a proof-of-concept Perturb-seq dataset targeting 100 genes across 16 diverse cancer cell lines, address single-cell technical artifacts, assess Cas9-associated chromosomal aberrations, and identify both common essential-gene signatures and context-specific responses, including an oxidative-stress-linked IER3IP1 dependency.
Why it matters: Parallel perturbation profiling across diverse cancer contexts provides a framework for separating conserved gene functions from tissue- or subtype-specific phenotypes while exposing technical confounders.
Why for Yiru: It directly supports computational oncology and perturbation biology, especially the design of context-aware single-cell screens for hypothesis generation.
Computational Pathology and Spatial Microdosimetry Guide Radiopharmaceutical Selection for TROP2-Targeted Alpha versus Beta Radionuclide Drug Conjugates (RDCs)
bioRxiv (Cancer Biology) Published 2026-08-25 Preprint DOI:
spatial omics computational pathology drug discovery microdosimetry radiopharmaceuticals
Summary: An automated computational-pathology and spatial-microdosimetry pipeline reconstructs heterogeneous TROP2 expression from whole-tissue immunohistochemistry and simulates absorbed dose distributions for 177Lu beta and 225Ac alpha radionuclide drug conjugates across 14 specimens.
Why it matters: The study makes spatial expression and radiation range explicit in choosing between therapeutic modalities, illustrating how tissue measurements can inform treatment design rather than merely describe heterogeneity.
Why for Yiru: It is a concrete bridge between computational pathology, spatial modeling, and drug discovery, with methods transferable to image-aware therapy-response modeling; the small specimen count and simulated doses limit clinical claims.
Cross-disciplinary watchlist
Other Fields
Spatial vascular/BTB remodeling and malignant-state plasticity in glioblastoma
bioRxiv (Cancer Biology) Published 2026-08-22 Preprint DOI:
spatial omics oncology tumor microenvironment glioblastoma blood-tumor barrier
Summary: Donor-aware analyses of 38 histopathology-annotated spatial transcriptomic sections from six donors identify spatially partitioned vascular and blood-tumor-barrier-associated programs alongside malignant-state plasticity. External cohorts support regional remodeling, while the authors distinguish the cross-sectional Recognition-Priming-Gate framework from a validated temporal cascade.
Why it matters: The work shows why vascular state and malignant state should be modeled as spatially heterogeneous programs rather than a binary barrier phenotype, while explicitly bounding causal and permeability claims.
Why for Yiru: It is a direct fit for spatial oncology and tumor-microenvironment method development, particularly models that connect tissue architecture to malignant-state variation.
Proteomics of human cancer-associated T cells identifies regulators of T cell functionality
bioRxiv (Immunology) Published 2026-08-23 Preprint DOI:
computational immunology proteomics tumor microenvironment T cells multiomics
Summary: Matched proteomic and transcriptomic profiling of dysfunctional and bystander CD8+ tumor-infiltrating T cells from treatment-naive non-small-cell lung cancer finds widespread discordance, with 8% of quantified proteins changing only at the protein level. Genetic perturbation identifies CHD4 and FASN as cell-intrinsic regulators of T cell function.
Why it matters: The protein-specific signals and perturbation validation show that transcriptomics alone can miss regulators of immune-cell dysfunction and metabolic fitness.
Why for Yiru: This is a strong rationale for integrated omics in computational immunology and tumor-microenvironment modeling, especially when inferring T cell state from single modalities.
Integrative spatial profiling of 3D genome organization and gene expression in tissue
Cell Published 2026-08-25 Research article DOI:
spatial omics 3D genome gene regulation spatial Hi-C gene expression
Summary: Spatial Hi-C-RNA simultaneously maps genome architecture and gene expression from the same tissue, revealing how spatial genome organization shapes cellular identity, development, and tumor evolution beyond what transcriptomics alone can capture.
Why it matters: Jointly observing regulatory architecture and expression addresses a major limitation of treating spatial transcriptomes as disconnected from the 3D genome.
Why for Yiru: The modality is potentially high-impact for spatial omics and tumor evolution and offers a conceptual route to connect tissue context with regulatory architecture; the supplied record is concise, so full-text details are needed before stronger claims.