Cao et al. present stFormer, a spatial transcriptomic foundation model that integrates intracellular context with spatial ligand signaling. stFormer improves cell- and gene-level analyses compared with a single-cell foundation model and enables in silico perturbation to investigate spatial ligand signaling responses.

Holste, Dos Santos, et al. develop SurpHer, a genetically encoded ratiometric biosensor for imaging extracellular pH at the surface of living cells. Through modular screening, they develop a membrane-localized SEpHluorin-mKate2 sensor that reports dynamic pH changes across physiological ranges. Stable SurpHer expression enables longitudinal visualization of extracellular pH gradients in a microfluidic pH gradient.

Nakashiba et al. engineer a multi-recombinase reporter mouse line, ROSA26-6R, consolidating responsiveness to six recombinases into a single locus. This platform minimizes colony management effort while providing robust tdTomato labeling for antibody-free 3D whole-embryo imaging and fine neural structure visualization.

Chueh et al. present a self-supervised sinogram inpainting framework to correct missing-wedge artifacts in soft X-ray tomography. By recovering unrecorded projection angles without ground-truth data, the model effectively restores cellular morphology and improves the volumetric quantification of complex structures like malaria hemozoin crystals.

Schienstock et al. present a technical framework using surgical preparation of the mouse spleen, intravital 3-photon microscopy, and quantitative image analysis to interrogate in vivo leukocyte dynamics in the splenic white pulp. This method facilitates the investigation of cellular behaviors that underpin spleen biology in health and disease.

Garcia et al. develop intersectional viral-genetic approaches to access molecularly overlapping dopamine and non-dopaminergic neuron subtypes in the mouse midbrain. Their strategy enables selective targeting of dopaminergic populations and reveals a previously unrecognized non-dopaminergic cell type distributed within the substantia nigra and ventral tegmental area.

Haus et al. create a learned, embedding-based representation of cell position in developing embryos. Embeddings describe dynamic morphology at the single-cell level, enabling applications in identification and phenotyping.

Kirbizakis et al. develop HisTrader, a tool that pinpoints the short stretches of accessible DNA, where regulatory proteins bind, from widely available histone modification data. HisTrader enables the identification of the active transcription factors and prioritization of disease-associated genetic variants without additional experiments to map open chromatin.

Gene-expression-based drug screening usually requires at least 50–100 copies of the same clone or cells treated in the same way. Ma et al. describe a single-cell analysis platform that enables the detection of therapeutically relevant transcriptomic signatures at a single-replicate level. This supports highly multiplexed single-cell drug discovery.

Leonard et al. present LIBRA-seq for nanobody discovery (nbLIBRA-seq), a method for rapidly screening and discovering antigen-specific nanobodies in alpaca. The approach increases the throughput and efficiency of nanobody discovery by screening for multiple antigens in the same library. With a single blood sample from a vaccinated alpaca, hundreds to thousands of antigen-specific heavy-chain-expressing B cells can be recovered and analyzed for B cell receptor sequences.

Biological networks can involve interactions among more than two components, but detecting them from dynamics is difficult. Altshuler et al. introduce a recurrent network framework that compares models at matched capacity to infer whether observed trajectories are best explained by pairwise, three-body, or mixed interactions across biological systems.

Self-amplifying RNAs (saRNAs) have the potential for rapid, durable, non-integrating delivery of transgenes in human induced pluripotent stem cells. Della Santina et al. show that transcription factors and reporters expressed from saRNA are present for multiple weeks, supporting forward programming to neurons and tracking electrophysiological properties in 3D cardiac spheroids.

Kim et al. present a multi-camera array scanner (MCAS) that uses 48 parallel lenses to image organoids. The MCAS reduces acquisition time by 98% compared to commercial systems. The platform enables rapid, automated phenotyping of growth, drug response, and viral transduction in 2D and 3D neural cultures.

Zang et al. present highSpaClone, a computational framework for inferring copy number alterations (CNAs) and tumor subclones from high-resolution spatially resolved transcriptomics (SRT) data. By integrating gene expression with spatial information, highSpaClone enables the study of tumor evolution and uncovers early genomic changes in surrounding tissues.

Weier et al. demonstrate applicability of plunge-freezing for mammalian brain tissue vitrification in the presence of physiologically appropriate cryoprotectant-containing solutions. Subsequent fluorescence-guided cryoelectron tomography and cryo-volume electron microscopy imaging reveal unperturbed tissue morphology and present plunge-freezing as a viable vitrification approach for thicker brain specimens.

Uwizeye et al. present an AI framework that segments whole-cell 3D electron microscopy volumes accurately using little training data and ordinary computers. By replacing laborious manual tracing, it makes large-scale comparison of microbial cell architecture faster, more reproducible, and more accessible.

Sato et al. present scstruc, an R package for learning and evaluating directed gene regulatory networks from single-cell transcriptomic data. By combining Bayesian network structure learning, causal-structure metrics, reference-free validation, and marker arc analysis, scstruc supports method selection and interpretation of condition-associated regulatory changes.

Redekop et al. present MARBLE, a multimodal contrastive pretraining approach for histology and genetic biomarker alignment. Fine-tuning the image-only module, using a real-world cohort of MSK-IMPACT samples, the authors demonstrate data-efficient biomarker prediction.

MacKenzie et al. present ON-Time, a rapid metagenomic sequencing and analysis workflow that characterizes microbial composition of individual patient fecal samples in

Xiao et al. present Deepeptide, a deep learning-based pipeline for identifying oligopeptide leads without requiring indication-specific training data. By linking disease-related information, biological processes, and molecular functions, Deepeptide identifies candidate oligopeptides across multiple metabolism-related contexts, supporting broader AI-assisted peptide lead discovery.

To advance lymphatic vessel engineering, Lightsey at el. develop a microfluidic device-based approach that integrates physiologically relevant extracellular matrix cues with oscillatory shear stress. Using human lymphatic endothelial cells cultured on dopamine-modified hyaluronic acid or fibronectin, this approach reveals how the matrix-flow interactions drive mechanotransduction, collecting vessel specification, and lymphatic morphogenesis.

Dissociation-based viral tracing loses neuronal spatial data, while EM and in situ sequencing are costly. Zhu et al. present FISH-detectable CASS barcodes with rabies monosynaptic tracing to capture anatomically intact synaptic connections. Accessible on routine imaging systems, the platform reveals layer-specific visual circuit wiring and works with other spatial profiling assays.

Manzato et al. develop COAST, a method to physically register images from consecutive tissue sections, enabling reliable downstream multi-modal analysis. Applying COAST to Stereo-seq and MALDI-MSI, the authors map the transcriptional and metabolic changes associated with kidney regeneration in an ischemia-reperfusion injury mouse model.

Wang et al. introduce MicroNucML, an open-source workflow for automated micronuclei segmentation that combines deep learning with nuclei assignment to enable scalable studies of genome instability and DNA damage responses.

Chen and Coquelin et al. present a multiplex genome editing workflow capable of generating highly modified mice with complex traits within a single generation. By combining editing modalities, this approach enables up to eleven simultaneous genomic modifications to advance the scale of genetically modified mouse models.

To address challenges associated with single-stain controls for spectral flow cytometry, Zambidis et al. design a bispecific antibody, termed CaptureBody, that binds to all leukocytes and then captures fluorescent-conjugated antibodies. CaptureBodies enable cell-based controls independent of the original antibody’s epitope and produce unmixing comparable to directly labeled cells.

Butters et al. show that sequencing-based estimates of tumor DNA content can be incorrect in genomically complex cancers, distorting interpretation of genome doubling and cancer-gene copy number. Flow cytometry of archival FFPE tissues provides a practical experimental reference, improving genomic analysis and revealing hidden genome doubling in genomically “quiet” sarcomas.

Luo et al. present mpTCRai, a deep learning framework for predicting TCR-antigen specificity. By mapping sequence mutations to structural perturbations and energetic consequences, mpTCRai serves as a computational pre-filtering engine to narrow down vast mutational spaces, guiding and accelerating the rational design of TCRs.

Piras et al. present a statistical framework to infer lineage hierarchies and phenotypic transition probabilities in heterogeneous cell populations without tracking individual cells. Using multi-type Bienaymé-Galton-Watson branching processes, the method reconstructs growth, plasticity, drug-response mechanisms, and LGR5/KRT20 dynamics in colorectal cancer tumoroids from cell-state abundance data.

Focused ultrasound can target deep brain regions but predicted and delivered pressures often diverge. Dadgar-Kiani et al. introduce a practical calibration framework based on sparse measurements and optimized skull attenuation that reduces this mismatch, enabling safer personalized transcranial ultrasound.

Shipkovenska et al. present an improved method for profiling cell-type-specific secretomes in complex tissues without the use of viral vectors. The authors perform proof-of-principle validation by profiling the secretomes of two rare cell types with innate immune functions: liver Kupffer cells and airway microfold cells.

Meyer et al. present a microarrayed human intestinal organoid platform for longitudinal, single-organoid analysis of cytokine-induced epithelial injury and recovery. The study introduces IORS to quantify epithelial restitution and applies the platform to benchmark bioactives that support barrier recovery under controlled inflammatory stress.

Lin et al. develop a color-size two-dimensional (2D)-encoded single-microbead imaging platform for the homogeneous analysis of four mycotoxins. By incorporating the YOLOv11 deep learning model to enable rapid data processing, the method offers a strategy for high-throughput food safety screening.

Wacker et al. present dcORBIT, a single-molecule fluorescence method based on DNA origami rotors that uses dye cycling to enable long-term measurement of DNA rotation at base-pair resolution. dcORBIT extends our toolkit to study molecular rotation and is here used to reveal molecular variability during transcription by E. coli RNAPs.

Badenoch et al. develop a machine learning and metabolic modeling framework that predicts how gut microbes alter colorectal cancer drug response. The model identifies unintuitive drug-microbe synergies and their mechanisms, generalizing across microbial species and treatment modalities, including immunotherapy, to enable microbiome-informed combination therapy.

Adachi et al. systematically benchmark AAV capsids in human cortical and thalamic organoids by integrating gene delivery, cell type tropism, cytotoxicity, and neuronal function. By identifying AAV-DJ as a low-toxicity, function-preserving vector, this study offers a practical platform for imaging, manipulating, and tracing human neural circuits in vitro.

Xiao et al. present SwitchClass, a label-switch classification framework that scores whether molecular features in intermediate states align with baseline or perturbation. Across proteomic, phosphoproteomic, and single-cell transcriptomic datasets, SwitchClass reveals recovered and persistent molecular programs after cancer treatment, dietary reversal, and COVID-19.

Brooks et al. report a human MSC-derived liver-on-a-chip integrating endothelial cells, hepatocytes, and cholangiocytes from a single donor source. The platform reconstructs the sinusoidal, hepatic, and biliary compartments, supports hepatic uptake and biliary excretion, and models ischemic liver injury with therapeutic cell recruitment for personalized therapy and liver research.

Takesue et al. enhance the nCas9-based gene amplification method BITREx through the development of (1) a plasmid reporter that predicts gRNA performance and (2) BITREx 2.0, a dual-nicking strategy that exploits synergistic interactions between two gRNAs. These advances broaden the utility of BITREx for gene amplification applications.

Chen et al. determined the pivotal role of sufficient axial resolution in maintaining data quality of volume electron microscopy despite varying image signal-to-noise ratios and developed a volumetric denoising algorithm. Volumetric denoising enables finer serial block-face cutting and improves organelle and synapse segmentation accuracy.

Peripheral innervation of adipose tissue is essential for metabolic regulation, yet gene delivery tools remain underdeveloped. Willows et al. establish a framework for targeting AAV-mediated transgene delivery to nerve subtypes in adipose tissue with Cre drivers, offering a promising tool for mechanistic studies and laying a framework for future therapeutic strategies targeting neuropathic disorders.

Hernaiz-Llorens et al. establish an AAV1-hSyn-TurboRFP workflow for streamlined corticospinal tract sprouting analysis. The method enables direct visualization of fine CST axons without secondary immunostaining, supports combined tracing and genetic manipulation, and provides reproducible quantification of injury-induced corticospinal remodeling.

Koutsoumpa et al. present OPTO-Cln2, an optogenetic system for light-controlled CLN2 expression in budding yeast based on an EL222-responsive promoter. In G1 cyclin mutant backgrounds, OPTO-Cln2 enables dynamic modulation of G1 progression and light-induced synchronization of cell populations.

Marin et al. evaluate ATP-Red as a simple flow-cytometry-based method for measuring ATP-associated bioenergetic states in single cells. The study shows that ATP-Red captures metabolic heterogeneity, immune cell activation, and mitochondrial dysfunction, providing a practical approach for single-cell metabolic profiling in complex biological systems.

Shaikh et al. develop a defined, xeno-free, scaffold-free human iPSC-derived prostate organoid platform supporting epithelial and stromal differentiation. Organoids generated through monoculture or modular co-culture approaches enable compartment-specific genetic manipulation and recapitulate key features of prostate tissue organization, providing a reproducible platform for developmental and disease studies.

Danning et al. introduce a feature selection method for single-cell expression analysis that combines statistical and ensemble learning techniques to identify genes that vary across the latent cellular trajectory. This preprocessing method improves downstream tasks including trajectory reconstruction, low-dimensional embedding, and pseudotime analysis.

Accurate influenza-like illness forecasting is essential for timely public health response but is often limited by scarce and privacy-sensitive surveillance data. Li et al. show that lightweight fine-tuning adapts pre-trained large language models to provide accurate and stable forecasts across diverse real-world surveillance datasets.

Rohner et al. develop Alpha1DLight, an α1D-adrenergic receptor-derived fluorescent indicator that enables subtype-resolved norepinephrine imaging. The sensor shows high affinity, reversible responses, and α1D-like pharmacological sensitivity and reports norepinephrine dynamics across cells, brain slices, and behaving mice, enabling investigation of sustained noradrenergic engagement and drug effects.

Understanding how proteins organize inside cells requires combining imaging and interaction data, yet doing so at scale remains technically demanding. Qian et al. present HIT-MAP, a streamlined pipeline that tags proteins at native genomic sites and acquires both modalities from one cell line, broadening access to multimodal cell mapping.