Nature | Methods Note

Nature | Methods

Nature Methods offers a unique interdisciplinary forum for the publication of novel methods. Nature Methods focuses on the life sciences, combining practical, technique-driven subject matter with rigorous peer-review standards to ensure that readers are consistently presented with only the most valuable and highest quality methodological research. The journal offers its readers primary research papers as well as an array of opinions, reviews and short journalistic pieces to provide busy researchers with a broad, yet easily absorbed perspective of important methodological developments in the life sciences.

Thread Of Notes

An image-based framework for the integrated analysis of the epithelial-to-mesenchymal transition

A quantitative imaging platform enables integrated analysis of the cellular dynamics associated with the epithelial-to-mesenchymal transition (EMT) in human induced pluripotent stem cells. Differences in EMT progression across 2D and 3D culture geometries were linked to basement membrane integrity.

A foundation model to help understand the regulatory implications of 3D genome organization

We developed a foundation model for 3D genome organization from Hi-C-derived chromatin structure. HiCFoundation supports integrative analysis linking genome architecture to downstream regulatory function. Through large-scale self-supervised pretraining on Hi-C data, followed by fine-tuning on downstream tasks, it provides a unified, efficient, generalizable and interpretable paradigm for 3D genome, single-cell and multiomics analysis across species.

Simultaneous two- and three-photon multiplane imaging across cortical layers in freely moving mice

A lightweight head-mounted multiplane microscope allows simultaneous imaging from >1,800 neurons spread across the cortical layers in freely moving mice performing complex behavioral tasks, sampled over weeks.

Evaluating the role of pretraining dataset size and diversity on single-cell foundation model performance

The performance of single-cell foundation models is dependent on many factors. This study assesses the effect of the pretraining dataset’s size and diversity, revealing potential challenges in pursuing consistent improvement by naively scaling up pretraining data.

Scaling up training dataset size for transcriptomic AI models is much pain with little gain

The advantages of training foundation models for single-cell datasets on large (tens of millions of cells) datasets have not been systematically tested. We evaluated the role of the size and diversity of the training dataset in the performance of single-cell foundation models and found little gain in increasing dataset size beyond a set point.

A scalable approach to investigating sequence-to-function predictions from personal genomes

Modeling the effect of DNA sequence variation on phenotypes such as gene expression faces unique challenges when deciphering inter-individual variation. This study presents a scalable and efficient sequence-to-function modeling framework for personal genome analysis.

A fast open-source wave optics simulator

We built an open-source library of wave optics models — Chromatix — that enables optics simulations to be scaled efficiently on modern computing hardware. Chromatix solves computational optics problems up to 22× faster than typical research code and enables researchers to share innovations as part of a standard library for wave optics models.

Brightness demixing for simultaneous multi-target imaging in 3D single-molecule localization microscopy

Brightness demixing is a method for simultaneously detecting different fluorophores in single-molecule localization microscopy based on their brightness rather than their spectral properties. As this approach allows imaging in a single channel, it simplifies and speeds up the imaging of multiple targets.

Neuropixels Opto: combining high-resolution electrophysiology and optogenetics

Neuropixels Opto probes combine high-density recording sites with optical waveguide-based light emission sites for simultaneous recording of neuronal activity and optogenetic stimulation with blue and/or red light. As demonstrated in the mouse brain, these probes allow the activation or silencing of defined groups of neurons while acquiring high-quality recordings.

Reading and writing neural activity with Neuropixels Opto probes

High-density electrophysiology devices allow neuroscientists to observe spikes from large populations of neurons, and optogenetics allows them to drive or suppress those spikes. We show that a single device can combine these two capabilities, providing a high-resolution means to both read and write neural activity in the living brain.

TADShop: systematic benchmarking and identification of topologically associating domains

This study benchmarks 43 computational methods for identifying topologically associating domains (TADs), presents a tool based on a consensus strategy that leads to improved performance, and releases a web service to benchmark and use TAD calling methods.

Adding chemical identity to cryo-electron microscopy

A correlative workflow combining cryo-electron microscopy and cryo-electron tomography with secondary ion mass spectrometry generates spatially registered chemical maps from the same vitrified sample, directly linking ultrastructural information with molecular composition. This advance opens new avenues for identifying where drugs, pollutants and signaling molecules reside within cells at nanoscale resolution.

AreTomoLive: automated reconstruction of comprehensively corrected and denoised cryo-electron tomograms in real time and at high throughput

AreTomoLive is an accelerated preprocessing pipeline for cryo-electron tomography that streamlines tomographic alignment, reconstruction and contrast enhancement. This pipeline prioritizes automation and throughput to deliver comprehensively corrected and denoised tomograms, including during data collection at scale.

A multimodal adaptive optical microscope for in vivo imaging from molecules to organisms

MOSAIC is a versatile multimodal microscope that allows seamless transition between various light-sheet, two-photon, label-free and super-resolution modalities. It is demonstrated in various systems ranging from cell culture to mice.

Recommendations and considerations for hydroxyl radical protein footprinting–mass spectrometry

This Perspective provides minimum community standards and best-practice guidelines for hydroxyl radical protein footprinting–mass spectrometry, including experimental design, sample preparation and oxidation, processing of oxidized samples, and analysis and interpretation of the resulting data.

Unraveling lncRNA diversity at a single cell resolution and in a spatial context across different cancer types

This resource leverages spatial and single-cell transcriptomics to investigate lncRNAs in cancer and reports 94,795 previously unannotated lncRNAs and new approaches to infer their functions.

Automated device for permitting free movement during simultaneous photometry and electrophysiology in mice

A device for simultaneous photometry and electrophysiology allows recordings in freely moving mice without restriction or inhibition of movement or entanglement of the fiber and cables.

FILM: mapping organellar metabolism by mid-infrared photothermal-modulated fluorescence

FILM is a mid-infrared photothermal microscopy variant that involves optical boxcar demodulation-based illumination, denoising and spectral deconvolution. Its relatively gentle nature allows imaging of metabolic processes in organelles in cell culture as well as in Caenorhabditis elegans.

‘FILMing’ the metabolic landscape of individual cell organelles

Optical boxcar-enhanced fluorescence-detected mid-infrared photothermal microscopy (FILM), together with artificial intelligence-assisted data processing, maps the chemical composition of individual cell organelles in their native context. This technique uncovered metabolic heterogeneity of lysosomes and lysosomal changes during physiological aging and pathological conditions.

Unsupervised transfer learning enables multi-animal tracking without training annotation

UDMT is a multi-animal tracker for behavioral research that is based on a transformer architecture, which negates the need for manually annotated training data. UDMT is showcased on datasets encompassing mice, rats, Drosophila, C. elegans and betta fish.