Research Radar — 2026-09-30
Methods & AI
Computational
A biological-response compound representation allows chemical perturbation prediction across cell lines
bioRxiv Subject Collection: Bioinformatics Published 2026-09-29 Preprint; primary RSS abstract reviewed, full manuscript not assessed DOI: 10.64898/2026.09.28.755146
perturbation prediction cellular representations drug discovery
Summary: BioPert represents a compound by its measured transcriptional effect in a reference cell line, then learns to predict expression changes in other cell lines. The authors report advantages over structure-derived representations and directly copying the reference response on Tahoe-100M and LINCS L1000, while noting batch-related variation in performance.
Why it matters: It makes a testable representation-design point: biological response measurements can provide information that chemical structure alone misses. The reference-copying control is particularly useful for distinguishing context transfer from a generic compound signature.
Why for Yiru: A useful baseline and evaluation design for virtual-cell and perturbation-response research, especially when assessing whether a representation predicts genuinely context-dependent responses.
Deciphering the gene-centric spatial dynamics across tissue and time with GeneSPOT
bioRxiv Subject Collection: Bioinformatics Published 2026-09-29 Preprint; primary RSS abstract reviewed, full manuscript not assessed DOI: 10.64898/2026.09.24.753111
spatial omics gene-centric analysis multiomics integration
Summary: GeneSPOT analyzes spatial relationships among genes and how those relationships reorganize across conditions. Its abstract describes tissue-structure recovery, paired and unpaired spatial multiomics integration, developmental program convergence, and aging-associated reorganization that would be missed by testing expression levels alone.
Why it matters: The gene-centric perspective complements cell-state clustering: two conditions can have similar gene abundance yet different tissue-level molecular organization. This offers a concrete alternative way to compare spatial programs.
Why for Yiru: Relevant to spatial-omics representation learning and comparisons of tissue organization across perturbations or disease states.
Platform-specific count-matrix preprocessing workflows for spatial transcriptomics data analysis
bioRxiv Subject Collection: Bioinformatics Published 2026-09-29 Preprint; primary RSS abstract reviewed, full manuscript not assessed DOI: 10.64898/2026.09.24.753981
spatial transcriptomics preprocessing benchmarking
Summary: This benchmark compares 371 count-matrix preprocessing workflows across 45 datasets from 11 spatial-transcriptomics platforms. The authors report that no workflow is best everywhere, while favorable choices are more consistent within platforms; they also examine segmentation and post-segmentation transcript processing for imaging-based assays.
Why it matters: Preprocessing can change the biological patterns available to every downstream model. A platform-aware benchmark is actionable infrastructure for interpreting apparent gains in spatial clustering, integration and representation learning.
Why for Yiru: Useful when building or comparing spatial-omics pipelines and deciding whether downstream results are robust to normalization, feature selection and segmentation choices.
Biomedical discoveries
Biomedicine
Spatially resolved chromatin architectures in mammalian brain tissues
Nature Methods Published 2026-09-25 Journal Article; abstract and indexed publisher excerpts reviewed DOI: 10.1038/s41592-026-03218-3
spatial epigenomics chromatin architecture assay development
Summary: Spatial Hi-C profiles chromosome contacts within tissue sections while retaining spatial information. The study reports region- and cell-type-associated chromatin architectures in adult and developing mouse brain, including single-spot reconstructions at 10-micrometre spatial resolution.
Why it matters: The assay adds three-dimensional genome organization to spatial measurements, providing a way to study regulatory architecture in anatomical context rather than only in dissociated populations.
Why for Yiru: A useful measurement framework for integrating spatial cell states with chromatin organization and for defining what spatial representation models could be tested against.
Photolabile oligonucleotides with topological light gradients enable spatially resolved single-cell transcriptomics and epigenomics
Nature Biotechnology Published 2026-09-25 Brief Communication; journal version of a 2025 preprint; abstract and indexed publisher excerpts reviewed DOI: 10.1038/s41587-026-03328-5
spatial single-cell multiomics spatial barcoding epigenomics
Summary: scSTAMP-seq uses photolabile oligonucleotides and spatially patterned light to encode cell positions, allowing spatial information to accompany single-cell transcriptomic and epigenomic measurements. The reported demonstrations include human embryoid models.
Why it matters: Light-encoded spatial indexing offers a modular route for connecting molecular cell states to tissue organization while using single-cell assay workflows.
Why for Yiru: Relevant to spatial multiomics and to designing datasets that connect cell-state representations with measured spatial context.
Senescence-directed nanotherapy ameliorates fibrosis and overcomes immune exclusion in cancer | Science
Science Published 2026-09-24 Journal Article; journal abstract and corresponding preprint excerpts reviewed DOI: 10.1126/science.aeg4791
spatial immune niches senescent-like macrophages cancer immunotherapy
Summary: Fucoidan-based nanoparticles target P-selectin-positive, senescent-like cellular states in fibrotic tissues and tumors. Single-cell and spatial analyses identify an immunosuppressive macrophage population; treatment remodels the surrounding niche, restores immune infiltration and improves checkpoint-therapy responses in preclinical tumor models.
Why it matters: The study links cellular-state identification to a targeted intervention and functional immune effects, rather than inferring therapeutic value from a disease signature alone.
Why for Yiru: Useful for spatial computational immunology and for designing perturbation studies that test whether changing a cell population alters its local immune environment.
RNF43 p.G659fs leads to natural killer cell dysfunction in MSI-high colorectal cancer through PI3K/AKT/mTOR signaling and HLA-E up-regulation | Science Advances
AAAS: Science Advances: Table of Contents Published 2026-09-23 Research Article; publisher PDF reviewed via an institutional repository DOI: 10.1126/sciadv.adz6907
spatial immunology single-cell analysis cell-cell communication colorectal cancer functional perturbation
Summary: Single-cell profiling and multiplex imaging link RNF43 p.G659fs in MSI-high colorectal cancer to an NK-cell-rich but functionally suppressed microenvironment. Isogenic tumor lines and patient-derived organoid cocultures support a PI3K/AKT-dependent increase in HLA-E; pathway inhibition restores NK-cell killing. Spatial imaging places inhibitory NKG2A-positive NK cells near HLA-E-positive tumor cells.
Why it matters: It connects tumor genotype, spatial immune organization and functional perturbation, showing why more immune-cell infiltration need not mean stronger antitumor activity. The proposed therapeutic axis is mechanistically supported but remains preclinical.
Why for Yiru: A useful template for genotype-aware spatial and single-cell studies that test cell-cell communication hypotheses with orthogonal imaging and functional rescue, rather than treating expression or proximity alone as causal evidence.
Targeting ZMYND8 unleashes IL-2 signalling to override T cell exhaustion
Nature Published 2026-09-23 Journal Article; primary publisher abstract and indexed results reviewed DOI: 10.1038/s41586-026-11059-5
single-cell CRISPR T-cell exhaustion cancer immunology
Summary: An in vivo single-cell CRISPR screen identifies ZMYND8 as a regulator of exhausted CD8 T-cell states. Follow-up experiments link ZMYND8 to suppression of IL-2 receptor–STAT5 signaling through p300-dependent Il2ra regulation, with genetic deletion improving T-cell function in preclinical infection and tumor models.
Why it matters: The study connects a cell-state perturbation screen to molecular and functional follow-up, a stronger evidentiary chain than assigning mechanisms from transcriptomic associations alone.
Why for Yiru: A concrete example for computational immunology and perturbation modeling: evaluate whether predicted state changes align with experimentally tested mechanisms and functional immune outcomes.