Reviews & Analysis

Anatomical brain mapping techniques face limitations in accurate fiber reconstruction. The cytoarchitecture-based link estimation (CABLE) framework enables 3D mapping of brain fiber tracts at cellular resolution without fiber-specific staining. Based on intrinsic cytoarchitectonic organization, CABLE provides a new approach for studying brain connectivity in humans and other primates.

Modeling cellular responses to developmental and chemical cues is essential for understanding disease progression and informing therapeutic strategies, yet it often demands extensive experimental screening. We have developed a conditional diffusion model called Squidiff, which enables the in silico prediction of single-cell transcriptomic responses to both developmental signals and perturbations.

ESPRESSO (Environmental Sensors Phenotyping Relayed by Subcellular Structures and Organelles) combines live-cell fluorescent dyes, hyperspectral microscopy, image processing and data analysis to relay high-dimensional cell phenotypes, the evolution of which can be monitored at the single-cell level for up to 24 hours.

A recently proposed Hodge Laplacian model has advanced single-cell multimodal data analysis by providing highly reliable results for complex multi-branching trajectories.

EpiAgent, a transformer-based foundation model pretrained on approximately 5 million cells and over 35 billion tokens, has advanced single-cell epigenomics by encoding chromatin accessibility as ‘cell sentences’. Benefiting from this framework, EpiAgent achieved state-of-the-art performance in typical downstream tasks and enabled perturbation response prediction and in silico chromatin region knockouts.

Two new methodologies for automated cell tracking allow rapid analysis of cellular-scale behaviors across diverse samples with minimal errors.

A new method for structure prediction of protein complexes, named GRASP, integrates restraints on the potential structure derived from experimental data from diverse techniques. GRASP provides more accurate and reliable predictions than existing methods for challenging systems including antigen–antibody complexes, integrative structure modeling and in situ protein interactions.

Manual analysis of single-molecule time traces is slow and subjective. Now, a transformer-based foundation model — META-SiM —automates key analysis tasks across diverse datasets and enables rapid, systematic discovery of subtle single-molecule behaviors. Application of this approach reveals a previously undetected pre-mRNA splicing intermediate, highlighting its potential to streamline biological discovery.

A new approach sheds light on the biological features learned by protein language models, promising greater interpretability for unsupervised sequence learning.

We developed scEpi²-seq, a single-cell method which jointly profiles DNA methylation and histone modifications from the same DNA molecule. This multi-omic strategy reveals how chromatin context influences the maintenance of DNA methylation and identifies distinct epigenomic signatures across cell types in both cultured and primary tissues.

A cryogenic scanning transmission electron microscopy (STEM) approach for analyzing thick biological specimens expands the reach of cryo-electron microscopy.

Two deep-learning frameworks — GHIST and iSCALE — turn routine histology images into a rich molecular resource, and predict spatial gene expression at single-cell resolution (GHIST) and at super-resolution across large tissue sections (iSCALE), for scalable, data-driven tissue biology.

This Perspective discusses challenges associated with sharing annotated image datasets and offers specific guidance to improve the reuse of bioimages and annotations for AI applications.

This Perspective explores the challenges and future directions in the world of human tumor microbiome research.

Targeted deaminase-accessible chromatin sequencing (TDAC-seq) measures chromatin accessibility across long chromatin fibers at targeted loci using double-stranded DNA cytidine deaminases. When combined with pooled CRISPR mutational screening, TDAC-seq enables the high-throughput detection of changes in chromatin accessibility following CRISPR perturbations, allowing fine mapping of sequence–function relationships within endogenous cis-regulatory elements.

We created T-CellAnnoTator (TCAT), a computational method that helps to identify T cell subsets, activation states and functions. It does this using reproducible gene expression programs found across many disease contexts and tissues. TCAT outperforms conventional approaches for T cell subset prediction, is easy to use programmatically or through a website, and can be adapted for other cell types.

A lightweight miniature two-photon microscope features multi-wavelength excitation, correction of aberrations, and interchangeable objectives for scalable fields of view. It enables multicolor, deep-brain and scalable neural imaging in freely moving mice.

We present MAPIT-seq, a method that uses antibody-directed RNA editing to concurrently profile in situ RNA-binding protein (RBP)–RNA interactions and transcriptome-wide gene expression in limited input material, including single cells and frozen tissues. This dual-omic strategy streamlines mechanistic analyses of post-transcriptional regulation in dynamic biological processes and clinically relevant samples.

We present a cost-effective ultra-high-throughput cytometry-based framework for the detection of physical interactions between cells, along with the characterization of complex cellular landscapes. Application of our approach can offer a systems-level understanding of immunity and facilitate study of the kinetics, mode of action and personalized response prediction of immunotherapies.

ProDomino is a machine-learning model that efficiently predicts domain insertion sites in host proteins on the basis of amino acid sequence alone. The model enables the greatly accelerated design of functional multi-domain proteins, such as light-triggered or drug-triggered protein switches.