Reviews & Analysis

Stimulated Raman scattering (SRS) microscopy has the capability to simultaneously visualize the spatial distribution of different biomolecules, but it remains challenging to reach super-resolution. To achieve this goal, a deconvolution algorithm, A-PoD, was developed and combined with SRS microscopy, enabling examination of nanoscopic biomolecular distribution and subcellular metabolic activity in cells and tissues.

Communication between cells is crucial for coordinated cellular functions in multicellular organisms. We present an optimal transport theory-based tool to infer cell–cell communication networks, spatial signaling directions and downstream targets in multicellular systems from spatial gene expression data.

A deep learning approach called DeepPiCt facilitates segmentation and macromolecular identification in the cellular jungle of electron cryotomography data.

Alignment of single-cell proteomics data across platforms is difficult when different data sets contain limited shared features, as is typical in single-cell assays with antibody readouts. Therefore, we developed matching with partial overlap (MARIO) to enable confident and accurate matching for multimodal data integration and cross-species analysis.

Dimension reduction is a cornerstone of exploratory data analysis; however, traditional methods fail to preserve the spatial context of spatial genomics data. In this work, we develop a nonnegative spatial factorization (NSF) model that allows interpretable, parts-based decomposition of spatial single-cell count data. NSF allows label-free annotation of regions of interest in spatial genomics data and identifies genes and cells that can be used to define those regions.

We developed an advanced deep learning approach called local shape descriptors (LSDs) to enable analysis of large electron microscopy datasets with increased efficiency. This technique will speed processing of future petabyte-sized datasets and democratize connectomics research by enabling these analyses using modest computational infrastructure available to most laboratories.

During the first two years of postnatal development, the human brain undergoes rapid, pronounced changes in size, shape and content. Using high-resolution MRI, we constructed month-to-month atlases of infants 2 weeks to 2 years old, capturing key spatiotemporal traits of early brain development in terms of cortical geometries and tissue properties.

We engineered a 3D outer-blood-retina-barrier (3D-oBRB) with a fully polarized retinal pigment epithelium (RPE) monolayer on top of a Bruch’s membrane and a fenestrated choriocapillaris network. This 3D-oBRB tissue faithfully recapitulates RPE– choriocapillaris interactions, dry age-related macular degeneration (AMD) phenotypes (including sub-RPE drusen deposits and choriocapillaris degeneration) and the wet AMD phenotype of choriocapillaris neovascularization.

This Perspective discusses available software tools for lipidomics data analysis and provides a web-based Lipidomics Tools Guide to help guide the choice of these tools, organized by the major tasks for lipidomics research.

Localization Model Fit (LocMoFit) is a tool that enables fitting of super-resolution microscopy data to an arbitrary geometric model. The fit extracts quantitative parameters of individual cellular structures, which can be used to investigate dynamic and heterogenous protein assemblies and to create average protein distribution maps.

We trained DEDAL, an algorithm based on deep-learning language models, to generate pairwise alignments of protein sequences taking into account the sequence-specific context of amino acid substitutions or gaps. DEDAL improved the alignment correctness on remote homologs by up to threefold and the discrimination of remote homologs from evolutionarily unrelated sequences.

We developed a FAIR (findable, accessible, interoperable, reusable) framework for researchers to share spatially standardized brain models. TemplateFlow enables the implementation of more reliable data processing pipelines by maximizing the accessibility of such models. It equips neuroimaging researchers with a foundational tool to bridge gaps between populations and species in neuroscience research.

The ability to measure protein complexes in single cells is currently limited to a very small number of targets. Combining a proximity ligation assay with single-cell sequencing creates the ability to measure hundreds of extracellular protein complexes and thousands of mRNAs in individual cells.

To accelerate data acquisition for in situ cryo-electron tomography, we created a method that takes into consideration sample geometry for the robust prediction of sample movement while the microscope stage is tilted. This approach enabled the parallel collection of tens to hundreds of tilt series.

By modeling the probability of N⁶-methyladenosine (m6A) RNA modifications for individual reads from direct RNA sequencing, m6Anet achieves high classification accuracy and takes a step towards transcriptome-wide maps of m6A modifications at single-base, single-molecule resolution.

Multiplexing real-time single virus tracking with imaging paves the way for detailed information on virus–host interactions, offering a potential paradigm shift.

Hyperfolder yellow fluorescent protein (hfYFP) and its variants are fluorescent proteins with high chemical and thermal stability. They resist aggregation, withstand diverse chemical challenges and show promise in expansion and electron microscopies. The chloride resistance and uncanny stability in guanidinium of hfYFP enable fluorescence-guided protein purification under denaturing conditions.

Common cellular segmentation models based on machine learning perform suboptimally for test images that differ greatly from training images. Cellpose 2.0 allows biologists to quickly train state-of-the-art segmentation models on their own imaging data. This was previously only possible using large, annotated datasets and required expert machine learning knowledge.

Detecting rare-variant associations in the noncoding genome is challenging. We present a scalable, flexible and streamlined rare-variant association analysis framework for biobank-scale whole-genome sequencing data, including gene-centric and non-gene-centric analyses by incorporating multiple variant functional annotations using various coding and noncoding units, conditional analysis, result summary and visualization.

A combination of light-sheet fluorescence microscopy (LSFM) with structured illumination doubles resolving power over LSFM alone. We show a practical implementation using a single objective for illumination and fluorescence detection and demonstrate its use for rapid volumetric imaging.