Tethering a protein of interest to anti-CRISPR converts protein degradation into amplified, programmable transcriptional outputs through Cas-based transcription factors, with sensitivity to protein loss set by the anti-CRISPR-Cas ratio. The circuits can be multiplexed for high-throughput discovery of protein-degrading and -destabilizing molecules. The work appeared in Cell.
RIED achieves super-resolution imaging from reaction-excited luminescence, including electrochemiluminescence, chemiluminescence and bioluminescence, by pairing a spatiotemporal recording strategy with entropy-weighted correlation and deconvolution adapted to reaction-driven photon statistics. It resolves intracellular organelles at approximately 100 nm, images surface proteins with high sensitivity, and supports 41-h continuous super-resolution live-cell imaging of mitochondrial transfer dynamics, all without laser excitation.
DartUniFrac, a new phylogeny-aware microbiome distance algorithm with GPU acceleration, computes UniFrac distances up to three orders of magnitude faster than UniFrac and scales to millions of pairwise samples and billions of taxa. It connects UniFrac with weighted Jaccard similarity and uses sketching algorithms for fast computation, removing the scalability bottleneck that has limited UniFrac on modern dataset sizes.
A recyclable lysosome-targeting chimera built from a polyzwitterion and a protein-of-interest ligand triggers macropinocytosis and degrades membrane or extracellular proteins without a dedicated receptor. The chimera dissociates in acidic endosomes, exits the cell via the endoplasmic reticulum-Golgi transcytosis pathway, and re-enters further degradation cycles. In a triple-negative breast cancer mouse model, these McR-TACs durably degraded programmed cell death ligand 1 or macrophage migration inhibitory factor and inhibited tumor growth.
Donor-complementary prime editing (DoPE) achieves precise, double-strand-break-free insertion of DNA sequences up to 12.5 kb in one step, using an odsDNA donor, a pair of opegRNAs and a PE2* prime editor, without recombinases or transposases. Short roughly 30-nucleotide overhangs support insertions from small fragments to over 10 kb, and oligonucleotide donor pools enabled in situ saturation mutagenesis of a targeted EGFP region. DoPE replaced PRKCSH mutant exons individually or simultaneously, correcting diverse mutations uniformly in vitro.
Prime assembly (PA) adapts prime editors to create annealing 3'-flaps on genome and donor DNA, enabling kilobase-scale insertion or megabase-scale excision at endogenous loci. It replaced endogenous sequences with a 2.9-kb donor at up to 57.8% efficiency in HEK293T cells with >90% accuracy, and integrated a chimeric antigen receptor cassette at up to 28.1% in primary human T cells. In mice, hydrodynamic delivery of a GFP donor yielded 4.3% integration in GFP-positive hepatocytes.
The paper introduces macropinocytosis-mediated recyclable LYTACs (McR-TACs), a receptor-independent approach to degrading extracellular and membrane proteins. The design is meant to exploit macropinocytosis so that the degrader is recycled and can act on its target across multiple cycles, addressing the reliance of conventional lysosome-targeting chimeras on specific cell-surface receptors. The title positions the work within the extracellular targeted protein degradation field.
A blinded, prospective benchmark (AIntibody) tested 511 AI-designed or predicted antibodies from 29 organizations across three tasks: affinity maturation, affinity ranking within HCDR3 clusters, and CDR design outside selection outputs. Several groups produced developable antibodies with affinities <100 pM, but successes did not transfer across tasks; affinity ranking was worse than random clone picking except for one model, and out-of-library design often failed to beat standard selections. The results show AI can optimize antibodies in defined, biologically grounded regimes while exposing gaps in affinity prediction, library-inspired design, and cross-task generalization.
SpotMAX is a generalist AI-assisted framework for automated detection and quantification of spot-like structures in microscopy images, working on 3D data with a graphical user interface plus cell segmentation and tracking. On a large 3D dataset it outperforms or matches state-of-the-art tools and expert human annotators, and it delivered new biological insights in applications from meiotic crossovers in C. elegans to mitochondrial DNA dynamics in yeast and telomere length in mouse stem cells. It replaces application-specific tools and manual annotation with a single multidimensional workflow.
A Cell paper, "Integrative spatial profiling of 3D genome organization and gene expression in tissue," presents Spatial Hi-C-RNA, a method that reads 3D genome organization and the whole transcriptome simultaneously in intact tissue in situ.
Why it matters: It reads 3D genome structure and the whole transcriptome in parallel in the same tissue, extending spatial transcriptomics and single-cell Hi-C methodologically.
A virus-like particle toolkit delivers CRISPR ribonucleoproteins to human monocytes, macrophages and dendritic cells, supporting gene knockout, base editing, epigenetic silencing and, with AAV donor delivery, site-specific integration by homology-directed repair. SLICeVLP, pairing VPX-lentivirus sgRNA delivery with VLP Cas9 delivery, enabled pooled loss-of-function and Perturb-seq screens in human macrophages. Screens converged on TNFAIP3 as a central regulator of inflammatory polarization; its ablation produced a proinflammatory state resistant to suppressive repolarization and enhanced CAR-macrophage cytotoxicity.
dsRNAscan, a new computational method, predicts double-stranded RNAs genome-wide without relying on A-to-I editing sites. It predicted 5 million dsRNAs in the human genome, with about 2.4 million high-confidence predictions identified using machine learning trained on editing and RNA structure-probing data. The work also predicts hundreds of vertebrate-conserved dsRNAs, editing-enriched regions suggesting intermolecular dsRNAs from sense-antisense transcripts, and shows the ratio of cytoplasm-accessible intramolecular to intermolecular dsRNAs correlates with ADAR dependency across cancer cell lines.
Prime assembly (PA) uses prime editing to create complementary 3' single-stranded DNA flaps on genome and donor, enabling double-strand-break-free megabase-scale genomic excision and multikilobase donor insertion in human cells. The system achieved 57.8% replacement efficiency at endogenous loci with multikilobase payloads, and in primary human T cells it integrated a chimeric antigen receptor site-specifically with up to 28.1% efficiency. It offers precise simultaneous large-scale deletion and insertion where existing DSB-based and recombinase-based methods fall short.
Why it matters: It enables double-strand-break-free kilobase-scale gene replacement in human cells, including primary T cells, with efficiencies no existing system reaches.
DartUniFrac is a new UniFrac algorithm that scales beta-diversity computation to millions of samples with billions of taxa, where existing UniFrac implementations become a bottleneck because pairwise comparisons grow quadratically with sample number. The authors prove that unweighted and weighted UniFrac are forms of weighted Jaccard similarity on tree branches, and use a succinct balanced-parentheses tree representation with rank/select and parenthesis-matching indexes for efficient traversal of billion-taxon trees.
aDISCO is a tissue-clearing method that enables 3D microscopy, clearing and immunolabeling of large archival FFPE human tissue blocks stored for 15 years or more. It worked across human brain, spinal cord, peripheral nerve, skin, muscle, heart, kidney, liver, spleen, colon, and lung with a broad range of antibodies. Applied to focal cortical dysplasia with deep learning-based analysis, it revealed disrupted cortical layering and focal and global neuronal density variations that conventional histology would likely miss.
Spatial Hi-C-RNA maps genome-wide chromatin contacts and transcriptomes from the same tissue section at near-single-cell resolution. Applied to mouse brain, developing embryos, and human melanoma, it tied A/B compartments, topologically associating domains, and loops to region- and cell-type-specific transcriptional programs, resolved chromatin and transcriptional remodeling during embryonic neuronal maturation, and identified melanoma intratumoral subdomains that RNA alone did not detect. It adds in situ 3D genome structure as a measurable layer of tissue biology.
Systematic affinity purification-mass spectrometry mapping of protein-protein interactions for 100 high-confidence autism spectrum disorder genes uncovered more than 1800 interactions. Pathogenic missense mutations in different genes converge on shared protein complexes and drive convergent rewiring of these interactions. Distinct patient-derived FOXP1 variants disrupt interactions with FOXP4 and alter cortical neurogenesis and neural activity in brain organoids, linking genetic variation to protein networks in ASD.
OneCell CUT&Tag delivers matched epigenome, full-transcriptome and surface marker profiles from every cell, with input as low as one cell and no need to aggregate cells computationally. Applied to mammary gland, it uncovered epigenomic priming of basal cells and showed that epigenomic and transcriptional remodeling are not fully synchronized during basal-to-luminal transdifferentiation. It also identified a role for H3K27me3 in shaping zygotic expression programs.
OptiPrime, a machine learning model grounded in prime editing mechanism, predicts PE efficiency and extends to PE3 and twinPE outcomes, outperforming prior models. It captures determinants of mammalian mismatch repair and nominates MMR-evasive silent edits that raise editing efficiency. In primary human and mouse cells and in a mouse model of KIF1A-associated neurological disorder, it guided efficient in vivo correction of a pathogenic brain mutation.
Molecular Self-Assembly Induced Cloning couples orthogonal in vitro self-assembly of overlapping DNA segments with host-cell DNA repair in vivo, enabling over 1,000 distinct gene fragments to be produced in one pot from microchip-synthesized oligonucleotides. The method achieves near-zero oligo misalignment, with synthesis errors held constant and controllable. Applied to PETase, it yielded massive variant libraries containing higher-potency variants than the gold standard enzyme.
