Research Radar — 2026-10-04
Methods & AI
Computational
FLIGHTED: Inferring fitness landscapes from noisy high-throughput experimental data.
Cell Systems (via PubMed) Published 2026-10-02 Peer-reviewed Cell Systems research article, online October 2, 2026; journal abstract/date review with explicitly identified 2024 preprint antecedent DOI: 10.1016/j.cels.2026.101739
Bayesian inference Protein fitness Experimental noise Model evaluation AI for science
Summary: FLIGHTED models noisy high-throughput protein assays with Bayesian inference to produce probabilistic fitness landscapes. Across selection and base-editing-linked assays, noise-aware labels improve downstream model performance and motivate rechecking model rankings and data-versus-model scaling conclusions.
Why it matters: Experimental measurement error can alter both the learned fitness landscape and the apparent merit of a machine-learning model; preprocessing and uncertainty are part of the benchmark.
Why for Yiru: Directly relevant to uncertainty-aware AI-for-science evaluation and to separating biological signal, assay noise and model capacity in perturbation-style datasets.
Regulon-informed cellular representations reveal task-dependent generalization in drug combination prediction
bioRxiv Subject Collection: Bioinformatics Published 2026-09-29 Preprint, not peer reviewed; RSS-abstract and official archive review DOI: 10.64898/2026.09.23.753774
regulons drug combinations domain shift evaluation
Summary: A DrugComb comparison tests landmark expression, pathway scores and regulon activity across cell-line, drug-scaffold and joint holdouts. Pathway-plus-regulon gains vary with the split and objective, and can reverse under simultaneous context and scaffold exclusion.
Why it matters: Shows that adding biological structure does not uniformly improve transfer or top-candidate prioritization.
Why for Yiru: A concrete evaluation template for cellular representations, drug combinations and observed-response models.
Single-allele nanoscale mapping of regulatory variants
Nature Genetics Published 2026-10-02 Peer-reviewed regulatory-genomics methods research; targeted primary-text review DOI: 10.1038/s41588-026-02776-8
variant-to-function allele-specific chromatin immune regulation causal perturbation
Summary: MCCv measures allele-specific nanoscale chromatin structure and enhancer–promoter contacts, linking regulatory variants to phased expression. Applied to 405 immune-disease-associated cis-regulatory elements, it identifies a neo-CTCF mechanism that interrupts contacts with SESN3 and follows it with editing and mouse evidence.
Why it matters: It creates a testable route from noncoding association to molecular mechanism rather than assigning targets by genomic proximity alone.
Why for Yiru: A strong template for combining multiomic evidence with perturbations when evaluating immune-cell regulatory programs.
Near-atomistic simulations reveal the molecular principles that control chromatin structure and phase separation
Nature Communications Published 2026-10-03 Peer-reviewed computational biophysics research; abstract and preprint-lineage review DOI: 10.1038/s41467-026-78050-6
chromatin organization multiscale modeling phase separation histone acetylation
Summary: OpenCGChromatin is a near-atomistic coarse-grained model linking DNA spacing and histone-tail chemistry to chromatin folding and phase separation. Predictions are compared with cryo-ET structures and biochemical condensate measurements; 108-nucleosome simulations probe acetylation patterns.
Why it matters: It makes specific molecular interactions interpretable while extending the size of simulated chromatin systems.
Why for Yiru: A useful mechanistic counterpart to cell-state and chromatin models that infer relationships primarily from omics data.
SVPG: a pangenome-based structural variant detection approach and rapid augmentation of pangenome graphs with new samples
Nature Methods Published 2026-09-21 Peer-reviewed computational methods article DOI: 10.1038/s41592-026-03219-2
pangenomes structural variants long-read sequencing benchmarking
Summary: SVPG uses long-read evidence and pangenome graphs to call structural variants and add new samples to a graph. The authors report strong cross-platform accuracy and nearly tenfold faster augmentation in a 20-sample benchmark.
Why it matters: Connects graph-based variant discovery with practical updating of growing pangenome references.
Why for Yiru: A useful genomics method and benchmark-design example for testing reference bias, rare-variant sensitivity and population transfer.
Biomedical discoveries
Biomedicine
Spatial transcriptomics reveals microenvironmental heterogeneity in osteoarthritic subchondral bone
Nature Communications Published 2026-10-02 Peer-reviewed Nature Communications accepted article in press; publisher abstract/date and targeted primary-manuscript review DOI: 10.1038/s41467-026-77518-9
Spatial transcriptomics Cell-cell communication Tissue microenvironment Osteoarthritis Inference limitations
Summary: Spatial profiling of five specimens from three hip-osteoarthritis patients maps an osteogenic core-halo organization and neighborhood changes in subchondral sclerosis. The analysis nominates FN1-SDC2 and COL1A1-DDR2 interactions and altered metabolic states. Visium HD was aggregated to 56-micrometer bins, so this is not a single-cell-resolution map.
Why it matters: The study combines tissue architecture, deconvolution and communication analysis in mineralized tissue, providing a reusable spatial-analysis resource with explicit inferential limits.
Why for Yiru: Useful for comparing spatial neighborhoods and cell-cell interaction models. The manuscript distinguishes measured spatial structure from inferred signaling and metabolic flux.
Systematic transcriptomic evaluation of blastoid models of early human development.
Cell Systems (via PubMed) Published 2026-10-02 Peer-reviewed Cell Systems research article, online October 2, 2026; NLM abstract/date and author-resource review, not journal full-text review DOI: 10.1016/j.cels.2026.101738
Single-cell reference maps Developmental benchmarking Embryo models Annotation uncertainty
Summary: An integrated single-cell reference of human embryo development is used to compare lineage coverage, identity and developmental progression in stem-cell-derived blastoid models. The authors release processed reference and benchmarking data with an application for evaluating query datasets.
Why it matters: Provides a common molecular yardstick for deciding which parts of early development an embryo model reproduces and where it remains incomplete.
Why for Yiru: A useful example of reference construction, annotation uncertainty and context-aware benchmarking for single-cell models and developmental trajectories.
RNA splicing factor mutations drive myeloid neoplasm oncogenesis through protein complex poisoning
bioRxiv Subject Collection: Cancer Biology Published 2026-09-30 bioRxiv preprint; not peer reviewed DOI: 10.64898/2026.09.29.754980
single-cell multiomics proteomics RNA splicing myeloid neoplasms
Summary: In myeloid neoplasms, long/short-read single-cell profiling and low-cell proteomics reveal mutation-dependent decoupling of RNA and protein dynamics. The authors propose that mis-splicing disrupts protein complexes and use iPSC hematopoietic differentiation to investigate candidate mechanisms.
Why it matters: Transcript-only models may miss consequential protein-level states, even when they capture RNA changes.
Why for Yiru: A useful test case for multimodal disease-state modeling and choosing validation layers beyond scRNA-seq.
Histone readers MLLT1 and MLLT3 concentrate AID to confer locus specificity
Nature Published 2026-09-30 Peer-reviewed mechanistic research article DOI: 10.1038/s41586-026-11087-1
B-cell biology chromatin AID mutagenesis perturbation and rescue
Summary: MLLT1/MLLT3 loss abolishes AID-dependent mutagenesis, while tethered AID rescue restores activity. The readers locally enrich AID at susceptible loci; condensates are a proposed contributor.
Why it matters: Separates enzyme occupancy from the local concentration needed for mutational activity.
Why for Yiru: A useful perturbation-and-rescue example linking chromatin context to B-cell diversification and lymphoma-relevant off-target mutation.
SEEKER: A genome-scale library-on-library screening platform for deciphering T cell recognition of antigen
bioRxiv Subject Collection: Immunology Published 2026-10-01 Functional immunology screening preprint; not peer reviewed; primary-abstract review only DOI: 10.64898/2026.09.25.753432
T-cell receptors antigen recognition functional screening autoimmunity
Summary: SEEKER combines a TCR–peptide-MHC library, clonal activation readouts and linked-sequence enrichment. The preprint reports approximately 100 million combinations per run and 53 validated receptor–peptide pairs from ankylosing-spondylitis T cells.
Why it matters: Functional recognition screens can supply evidence that binding predictions and repertoire similarity alone cannot provide.
Why for Yiru: An enabling platform for connecting immune clonotypes to antigens and building experimentally grounded T-cell specificity datasets.