Skip to content

#方法/工具

Today 0 items
Oct 15Thu
  1. biorxiv.org40

    Beyond the gene: contextualizing novel isoforms through pathway analysis in long-read single-cell RNA sequencing

    A new computational framework, Centroid-based Isoform Pathway Inference (CIPI), assigns biological context to the thousands of novel isoforms and genes uncovered by long-read single-cell RNA sequencing. Instead of relying on host-gene annotation or known protein domains, CIPI builds expression-based neighborhoods of reference genes and combines them with GO, KEGG, or Reactome pathway analyses to infer pathway-level roles for individual isoforms. The result is a hypothesis-generating method for prioritizing novel isoforms across the isonome.

  2. biorxiv.org49

    Pervasive expression of early embryonic programs in cancer has stage-dependent survival impact

    Early embryonic gene expression programs are widespread in human tumours, and transcriptional similarity to distinct preimplantation stages independently predicts overall survival beyond age, sex, mutational status and molecular subtype. Similarity to zygote and 2-cell embryo stages predicted improved survival, driven in kidney cancer by stromal and immune cells, while trophectoderm-associated programs in tumour cells predicted worse survival in breast cancer and melanoma and were linked to MYBL2 and FOXM1 dependencies.

  3. biorxiv.org44

    Mitigating genotype-driven bias improves atlas-scale single-cell DNA methylome profiling

    Masking variant positions before DNA methylation quantification reduces genotype-driven bias in pooled single-cell bisulfite sequencing, where C-to-T variants are misread as conversion events. In a human blood nuclei sciMET dataset from 10 healthy individuals (25,674 nuclei), the approach reduced donor-associated confounding during clustering and fewer spurious differentially methylated regions within the same cell type, while improving batch mixing and preserving biological variation relative to uncorrected workflows.

  4. nature.com72

    Structural variant detection across sequencing technologies

    HitSV, a read-length-agnostic structural variant (SV) calling approach, models how local genomic repetitiveness distorts alignment signatures and combines progressive clustering with haplotype-resolved local assembly for SV reconstruction. Applied to long-read, short-read and hybrid data, it achieved high yields across sequencing technologies, reduced long-read sequencing data requirements, and maximized SV discovery from short-read data. Applied to a large-scale Chinese cohort, it identified composite repeat architectures featuring tandem-mobile-element arrays as a pervasive, previously underrecognized feature of human genomes.

    Why it matters: It improves structural variant detection across short-read, long-read and hybrid data and reveals underrecognized composite repeat architecture in human genomes.

  5. Nat Biotechnol63

    Multimodal learning enables chat-based exploration of single-cell data.

    CellWhisperer links transcriptomes to text through a multimodal embedding trained by contrastive learning on 1 million RNA sequencing profiles with AI-curated descriptions, feeding a large language model that answers questions about cells and genes in natural-language chats. Benchmarked for zero-shot prediction of cell types and biological annotations, it supported discovery in a meta-analysis of human embryonic development and is integrated with the CELLxGENE browser for combined graphical and chat exploration of gene expression.

  6. biorxiv.org42

    Spatial Tumour Tissue Distribution Dynamics of a Bicycle Peptide-MMAE Toxin Conjugate Profiled by Targeted Mass Spectrometry

    Laser capture microdissection coupled to targeted LC-PRM mass spectrometry maps the spatial distribution of MMAE released from the NECTIN-4-binding Bicycle Drug Conjugate zelenectide pevedotin (BT8009) at sub-tissue resolution in mice. The method showed progressive accumulation and sustained retention of MMAE in tumour tissues over 24 h, while levels in plasma, liver and muscle declined after early exposure, and revealed heterogeneous MMAE distribution between tumour cell-rich and stroma-rich regions with distinct temporal patterns.

  7. nature.com71

    Artificial intelligence-enabled clinical phenotyping of human brown fat

    AI-enabled quantification of brown adipose tissue (BAT) activity in humans links higher BAT activity to a healthier metabolic profile and lower rates of psychiatric disease. The same approach identified existing pharmacologic agents that could be repurposed to activate BAT in humans. The work provides a scalable way to phenotype BAT and connect it to clinical outcomes and drug targets.

    Why it matters: It shows AI-based BAT phenotyping at scale, linking brown fat to metabolic and psychiatric outcomes and to repurposable drugs.

  8. nature.com80

    Computational design of metalloproteases

    RFdiffusion was used to design proteins de novo with water-activating zinc-based active sites positioned next to a peptide substrate binding pocket. The resulting proteases were highly active and programmable, with the authors pointing to potential applications in therapeutics, biotechnology and bioremediation.

    Why it matters: De novo design of active, programmable metalloproteases extends computational protein design to catalysis.

  9. nature.com80

    Modelling mouse embryogenesis from chemically induced totipotent stem cells

    A mouse embryo model built from chemically induced totipotent-like stem cells progresses through key developmental milestones, including gastrulation, embryonic axis formation, haematopoiesis and early organogenesis. The work extends embryo modelling to a starting cell type with totipotent-like properties. It provides a system for studying development from the earliest stages onward.

    Why it matters: It shows totipotent-like stem cells can generate an embryo model reaching organogenesis.

Oct 14Wed
  1. biorxiv.org54

    Towards standards for analysing structural variants from short-read whole genome sequencing data

    Only five of sixteen structural variant callers reliably produced deletions, insertions and duplications that perfectly matched truth sets (HG002 T2T-Q100, Platinum Pedigree) from 30X Illumina WGS data, and ensemble pipelines such as GATK-SV, MOPline and Parliament2 reduced perfect matching accuracy by up to 98% compared to their best constituent tools. This inaccuracy propagates into gnomAD-SV and likely affects TOPMed, All of Us and Biobank Japan. An intra-sample merging strategy achieved the highest F1-scores for deletions and insertions on HG002, recovering more perfect matches than the best ensemble pipeline tested.

  2. biorxiv.org42

    cit_gsa: gene set analysis for scRNA-seq with covariate adjustment

    cit_gsa is a one-step self-contained gene set differential expression analysis method for single-cell RNA-seq that supports more than two conditions and covariate adjustment, unlike existing two-step approaches. Benchmarks show better performance than state-of-the-art methods in complex scenarios, and real-data analysis confirms that covariate adjustment improves biological interpretation.

  3. Nat Med71

    MAGIC: an international network for evaluating generative artificial intelligence in global health.

    The authors establish MAGIC, an international network for evaluating generative artificial intelligence in global health. The network brings together collaborators across institutions in Asia, Europe, Africa and the Americas, including groups at Duke-NUS Medical School, Harvard, Tsinghua and the World Health Organization Regional Office for South-East Asia, with a focus on the ethics, regulation and standardization of AI for health.

    Why it matters: Sets out an international framework for evaluating generative AI in global health, relevant to clinicians and researchers deploying these tools.

  4. Nat Biotechnol82

    In vivo engineering of CAR natural killer cells for immunotherapy using ligand-directed lipid nanoparticles.

    Ligand-directed lipid nanoparticles (NtLNPs) carrying mRNA transfected up to ~80% of NK cells in vivo in mice with minimal off-target effects, generating CAR NK cells without ex vivo engineering. CD19-targeted CAR NKs completely depleted circulating B cells in healthy mice, CD3-targeted CAR NKs eliminated nearly all autoreactive T cells in autoimmune hepatitis, and NKG2A-targeting versions improved solid tumor efficacy. Human-optimized NtLNPs produced CAR NK cells that regressed two humanized hematologic malignancy models. The work replaces viral, ex vivo CAR manufacturing with a modular in vivo mRNA delivery platform.

    Why it matters: It shows nonviral in vivo CAR NK generation, a route around ex vivo viral engineering of cell therapies.

  5. biorxiv.org45

    A multimodal precision oncology platform integrating genomics,single-cell transcriptomics, and functional drug screening expandstherapeutic options in pediatric brain tumors

    A tiered precision oncology framework combining deep whole-genome sequencing, single-cell RNA sequencing, and ex vivo functional drug screening was applied to 17 children with malignant CNS tumors. Functional screening gave every patient at least one high-confidence targeted-therapy or chemotherapy nomination, single-cell RNA sequencing found an average of 37 druggable gene targets per patient, and high-confidence targeted-therapy recommendations rose fourfold. Functional classifications matched outcomes in 8 of 10 evaluable cases.

  6. biorxiv.org40

    Dysregulation of epithelial differentiation architecture underlies field cancerization in head and neck cancer

    AnnZarro is a point-and-click browser viewer that displays cell-by-cell and gene-by-gene matrices such as kernels and correlations linked to an AnnData object's cells × genes data. It reads only the matrix rows and columns each panel displays from zarr stores, so users can explore matrices larger than memory: on a laptop, a 100,000-cell part of Tahoe-100M opens in 4 s, and all 95.6 million cells draw in one plot in about 5 to 7 s. The tool lets experimental biologists explore data without code and computational biologists probe the relationships behind their analyses.

  7. biorxiv.org40

    Predicting resistance to synthetic lethal therapies in cancer

    Resistance to synthetic lethal cancer therapies can be predicted from molecular network features. A random forest model combining protein-protein interaction, co-expression and essentiality data, trained on CRISPR-Cas9 resistance screens across ten SL biomarker-target pairs, achieved a mean AUROC of 0.72 under leave-one-SL-pair-out validation (range 0.57-0.91) and independently predicted resistance to the MTAP-PRMT5 and ATM-ATR therapies now in trials. The work gives a computational way to anticipate resistance mechanisms before they emerge.

Oct 13Tue
  1. Nat Biotechnol40

    Ten years, tens of millions of cells.

    The Human Cell Atlas consortium, launched in October 2016 by around 100 researchers led by Aviv Regev and Sarah Teichmann to map the body's roughly 30 trillion cells, has grown into one of the largest collaborative efforts in biology. By 2024 it had released four integrated v1 atlases, including a Human Lung Cell Atlas with over 2 million cells from 486 individuals across 49 datasets. Its next stage, cataloguing billions of cells, will draw on spatial biology and AI advances.

  2. biorxiv.org53

    Benchmarked sequencing methods reveal hidden mosaic variation in the human brain

    Benchmarking four duplex sequencing methods against bulk and single-neuron genome sequencing in matched human brain tissue showed that mosaic variant recovery is strongly method-dependent, with limited caller-level concordance. The new RIVER tool integrates read-level evidence, expanding cross-method-supported candidates by roughly two orders of magnitude; orthogonal sequencing at ~10,000x validated about half of tested RIVER-rescued candidates. Recovered variants resolved broadly distributed and regionally restricted clones and traced lineage among individual neurons.

    Why it matters: A benchmarked method that expands usable mosaic variant markers changes how brain lineage and mosaicism can be studied.

  3. doi.org43

    Modelling nano-particle-PEG-NLS complexes for nucleus-targeted drug delivery: revisiting the multi-dynein nano-cargo transport on microtubules

    An energy-based Monte-Carlo simulation of spherical nanoparticles hauled by multiple dynein motors via flexible PEG linkers reproduces motility assay results: for a mean of 1 to 6 motors per nanoparticle, predicted velocities, run-distances and run-times fall within experimental 95% confidence intervals. A scan of linker contour length for a 20 nm radius cargo with six motors finds an optimum run-time and run-length at about 40 nm, despite reduced velocity. The model gives quantitative design rules for tuning PEG linker length and PEG-NLS density in nucleus-directed nanoparticle delivery.

  4. Nat Biotechnol72

    Discovery and engineering of retrons for precise genome editing.

    Engineered retron reverse transcriptases identified from metagenomic data produce single-stranded DNA donors in mammalian cells at efficiencies comparable to conventional oligodeoxynucleotide donors, but from a genetically encoded cassette. The editors work with Cas12a nuclease and Cas9 nickase, broadening targetable sites and avoiding double-stranded breaks. An all-RNA delivery strategy enables DNA-free editing in cells and vertebrate embryos, demonstrated by inserting a split GFP epitope tag for live-cell imaging.

    Why it matters: It turns retrons into genetically encoded donor systems for precise editing without exogenous DNA or double-strand breaks.

  5. Sci Adv61

    MIMAS: An open-source microlens-array miniature microscope for accessible large-field cortical imaging in freely behaving mice.

    An open-source head-mounted fluorescence microscope using a 4 × 5 microlens array as its sole imaging element achieves cellular-resolution imaging over an approximately 4 × 4.5 mm2 field of view at 2.31 g. In freely behaving mice, it captured calcium signals from thousands of layer 2/3 neurons across multiple dorsal cortical regions, plus wide-field vascular imaging with vessel caliber-dependent hemodynamic responses to isoflurane. It offers labs a low-complexity route to large-scale cortical imaging.

    Why it matters: Open-source single-array optics put cellular-resolution wide-field cortical imaging within reach of ordinary labs.

  6. biorxiv.org40

    Alignment-Aware Genomic Coordinate Encoding for Long-Range DNA Modeling

    USM-AA is a genomic foundation model that encodes reference genome coordinates through a non-contiguous positional encoding called Alignment-Aware Rotary Position Embedding, plus an alignment embedding capturing chromosome, strand, and CIGAR information. It uses reverse-complement masked learning to improve long-range sequence modeling and was pretrained on alignment-aware sequences spanning cross-species evolution and human population variation. USM-AA consistently outperforms existing models across multiple downstream tasks, supporting alignment-aware representations for modeling distal regulatory elements and sparse motifs.

  7. nature.com51

    Live-cell imaging goes WHOLISTIC

    WHOLISTIC is a live-imaging platform that records cellular activity across the entire vertebrate body at single-cell resolution, using genetically encoded calcium sensors, high-speed volumetric fluorescence imaging, computational motion correction and tissue-scale activity analysis. Ruetten and colleagues built transgenic zebrafish expressing a calcium indicator in nearly all cell types and demonstrated the approach in adult Danionella cerebrum, with whole-body expansion microscopy supporting molecular and anatomical characterization of cell populations.

Oct 12Mon
  1. nature.com48

    Private, offline and yours: how to search the literature with local AI systems

    Growing numbers of researchers are running large language models offline on their own machines, using tools such as Ollama and LM Studio to sort, query and index their literature libraries while keeping sensitive data out of cloud systems. A 2025 analysis found that six major AI developers trained models on users' input data by default, and a Wiley survey of some 2,400 researchers found 68% had used AI for literature searches or reviews. Local setups trade some capability, such as internet-based literature search, for privacy, reproducibility and free inference.

Oct 11Sun
  1. biorxiv.org40

    BoNTax: a curated genome-based framework for taxonomic classification and characterization of Clostridium botulinum Group I and Clostridium sporogenes

    BoNTax decouples ANI-based species assignment from botulinum neurotoxin characterization for Clostridium botulinum Group I and Clostridium sporogenes. Analysis of 966 genomes informed selection of 46 references that resolved into two species clusters with within-species ANI >95% and between-species ANI <95%, supporting a 95% species threshold. Of 42 validation genomes, 29 of 32 reported assignments were concordant, three prompted re-evaluation, and ten unknown genomes received ANI-based assignments. BoNT typing, including bivalent strains, was fully concordant.

    Why it matters: Gives public-health labs a standardized GalaxyTrakr workflow that separates species identity from toxin genotype in closely related clostridia.

  2. Sci Adv66

    Contagion-preserving compression for multiscale epidemic modeling.

    Iterative structural coarse-graining (ISCG) compresses large networks into multiscale representations that preserve contagion dynamics. In the saturation regime the compressed networks reproduce outbreak sizes, node-level infection risks, and spatiotemporal infection trajectories. Interventions designed on these representations, including influence maximization, immunization, and surveillance, consistently outperform adaptive centrality-based methods.

    Why it matters: Offers a compression method that preserves contagion dynamics and improves intervention design on large networks.

Oct 10Sat
  1. Cell76

    Mapping transcriptional responses to cellular perturbations with RNA fingerprinting.

    RNA fingerprinting maps transcriptional responses from new single-cell experiments onto reference perturbation dictionaries, assigning query cells probabilistically to one or more candidate perturbations. Benchmarks on ground-truth datasets show accurate single-cell assignments, scalability to genome-wide screens, and resolution of combinatorial perturbations. Applications include identifying p53 regulators under ribosomal stress, dose-dependent drug mechanisms of action, and cytokine-driven B cell heterogeneity in secondary influenza infection in vivo.

    Why it matters: It provides a framework for causal interpretation of single-cell data by matching cellular states to perturbations.

  2. biorxiv.org62

    Systematic mapping of enhancers, silencers, and topological elements with prime editing deletion screens

    Swap-seq, a method combining pooled twin prime editing with flow-sorting, deletes hundreds of endogenous regulatory sequences and measures their quantitative effects on nearby gene expression. Applied to PPIF regulation in THP-1 monocytes, it quantifies distal and intronic enhancers, characterizes CTCF insulators, and uncovers a silencer that represses neighboring enhancers through HDAC-mediated mechanisms, with hundreds of genomic elements sharing its signatures. The resulting data show sequence-to-function models capture local element activity but fail to link elements to distal target genes.

    Why it matters: It delivers a generalizable deletion-screening tool and a benchmark showing sequence-to-function models miss distal gene targeting.

  3. biorxiv.org51

    A three-dimensional imaging pipeline for the detection of vasculogenic mimicry in mouse tumor models

    A 3D light-sheet imaging pipeline detects functionally perfused vasculogenic mimicry in mouse tumors, overcoming artifacts of 2D CD31/PAS staining that cannot confirm vessel function. Mice were injected with fluorescent tomato lectin, and tumors were optically cleared and stained with endothelial markers. Applied to six breast cancer patient-derived tissue xenograft models, an orthotopic mammary cancer model and two small cell lung cancer circulating-tumor-cell explant models, it revealed VM vessels, mosaic vessels and host-VM junctions.

Oct 9Fri
  1. biorxiv.org66

    Random mtDNA mutagenesis maps genotype-phenotype associations across metabolic contexts and lineages

    mtDIVE, a random C→T mutagenesis platform, enables pooled mitochondrial forward genetics in human epithelial cells and zebrafish embryos, generating molecules from single substitutions to more than 100 mutations. Selection on mutant genomes and their transcriptional consequences depend on metabolic state and developmental lineage: obligate OXPHOS eliminates highly mutated cells by apoptosis, while glycolytic conditions promote their competitive displacement. Individual loci have modest, non-redundant effects that predict cell state beyond mutation burden, and cancer-recurrent variants are enriched among mutations linked to reduced mitochondrial transcription.

    Why it matters: It replaces one-allele-at-a-time mtDNA models with pooled forward genetics separating mutation identity from load.

  2. biorxiv.org60

    STEER-seq: a recombinase based platform for pooled multimodal single-cell perturbation screening

    STEER-seq is a virus-free platform for pooled ORF gain-of-function screening in human induced pluripotent stem cells, using Bxb1-mediated integration at a defined safe-harbour locus so each perturbation is expressed from a single copy in a common genomic context. Joint RNA and chromatin accessibility profiling linked transcription factor responses to chromatin remodelling. In a haematopoietic screen, SPI1 alone drove differentiation to HSC/MPP-like progenitors without exogenous cytokines, and scaling to around 2,000 human coding sequences uncovered factors inducing neural progenitor-, melanocyte- and vascular smooth muscle-like states.

    Why it matters: It removes random viral integration as a confound in pooled gain-of-function screens and works in hiPSCs at ~2,000-ORF scale.

  3. Sci Adv63

    Toward transgenic multicellular parasites: CRISPR-Cas12a germline editing of Schistosoma mansoni eggs as key to success.

    Cas12a outperformed Cas9 for homology-directed insertion of a 5'-C6-PEG10-modified donor template with 50-nucleotide microhomology arms into the GSH1 safe-harbor site of Schistosoma mansoni eggs. Electroporated eggs expressing EGFP under an S. mansoni ubiquitin promoter yielded transgenic cercariae that produced EGFP-positive adult worms in hamsters. Sequencing confirmed stable chromosomal integration transmitted through filial generations, opening functional genomics in platyhelminths.

  4. biorxiv.org41

    Mathematical Modelling of Stochastic Gene Expression Quantifies mRNA Heterogeneity in Cancer Cell Populations with Extrachromosomal DNA

    A stochastic model of gene transcription quantifies mRNA production and decay in cancer cells carrying ecDNA, exploring the space of mRNA distributions possible from a given ecDNA distribution. Applied to single-cell measurements of ecDNA and mRNA counts in cell lines, it estimates transcriptional parameters that are strikingly similar across cell lines and amplification types. Simulating genotoxic therapy, the model predicts the mRNA composition of persisting cells, linking ecDNA-driven heterogeneity to treatment resistance.

  5. Nat Biotechnol60

    A single-chain derivative of an integrin-activating antibody potentiates organoid growth in Matrigel and collagen hydrogels.

    A single-chain version of the integrin-activating antibody TS2/16 boosts growth of human gastrointestinal organoids, increasing yield up to fivefold in Matrigel and six- to sevenfold in collagen I hydrogels in both 3D and 2D. By pairing scTS2/16 with clinically available collagen I, the work offers a route away from Matrigel toward defined, clinical-grade culture conditions.

  6. doi.org40

    Innovative niosome screening for efficient gene delivery in senescent primary umbilical cord stroma-derived MSCs

    DP20CQ niosomes, built with DOTMA and polysorbate 20 plus chloroquine, delivered genes more effectively than DP80CH formulations in Palbociclib-induced senescent UC-MSCs from four donors. At DOTMA/DNA ratios of 10/1 and 15/1, DP20CQ achieved the highest transgene expression in senescent UC-MSCs, surpassing Lipofectamine, with strong endosomal escape at 10/1 and no loss of cell viability. This positions niosomes as gene delivery vehicles for senescent cells in antiaging therapy.

  7. nature.com47

    Phylogeny-informed genomic prediction with GPN-Star

    GPN-Star, a phylogeny-informed genomic language model, predicts functional genetic variants contributing to inherited traits by learning evolutionary constraints from multispecies whole-genome alignments and the corresponding phylogeny. Trained on vertebrate, mammalian and primate alignments, it predicts coding and non-coding variant effects on human disease susceptibility and enhances enrichment of complex trait heritability. Versions trained on five model organisms, from Mus musculus to Arabidopsis thaliana, show the approach generalizes across species.

Oct 8Thu
  1. Nat Biotechnol40

    High-throughput metabolic engineering in mammalian cells.

    Shotgun genetic engineering (SGE) enables multiplexed engineering of complex phenotypes in mammalian cells, screening millions of combinatorial solutions in a single experiment. Selection strategies reveal effective designs, allowing rapid engineering of complex biological functions. The method addresses the difficulty of programming multigene traits that has made mammalian metabolic engineering largely intractable.

  2. Nat Med46

    AI-assisted scoping review of code sharing in clinical prediction model research.

    Only 482 of 3,967 articles citing TRIPOD or TRIPOD+AI (12.2%) included code-sharing statements, according to a large-language-model-assisted scoping review that screened articles and assessed repositories against 14 reproducibility-related features. Sharing varied by journal and country, and repositories showed substantial heterogeneity. The results inform TRIPOD-Code, a reporting guideline for code availability and reproducibility.

Oct 7Wed
  1. Nat Biotechnol63

    Decoding systemic vascular health and hypertensive disorders in pregnancy through retinal imaging and Visionary AI.

    Visionary AI, a platform combining retinal imaging with graph-based vascular modeling, predicted hypertensive disorders of pregnancy before clinical onset. In a prospective development cohort of 1,267 pregnancies it reached AUC 0.91 and average precision 0.81 for preeclampsia, and a stability-optimized model transferred without retraining to external validation at AUC 0.81 and AP 0.68, outperforming clinical and deep learning benchmarks. The finding positions the maternal retina as a minimally invasive biosensor for first-trimester risk stratification.

    Why it matters: First-trimester preeclampsia prediction from retinal images outperformed clinical and deep learning benchmarks in external validation.

  2. Nat Biotechnol68

    Highly multiplexed mammalian metabolic engineering with a shotgun approach.

    Shotgun genetic engineering (SGE) delivers many barcoded small constructs into mammalian cells, so each cell tests a synthetic pathway varying gene content, stoichiometry and organellar localization, with functional pathways found by sequencing barcodes from phenotypic cells. Screening millions of pathway combinations enabled near-wild-type growth without valine and isoleucine-free growth in Chinese hamster ovary cells, plus valine-free growth in Jurkat cells. Functional pathways favored mitochondrial localization and involved tens of kilobases of synthetic DNA, and the datasets support machine-learning-guided engineering of complex biosynthetic traits.

    Why it matters: It makes unbiased combinatorial pathway screening in mammalian cells tractable at a scale conventional methods cannot reach.

CBACBA Frontier Insights (CFI)Chinese Biopharmaceutical Association (CBA)www.cba-usa.org