Software articles from across Nature Portfolio

Single-cell bisulfite sequencing enables the genome-wide quantification of DNA methylation at single-cell resolution, but methods to analyze the resulting data are lacking. The MethSCAn software accurately distinguishes cell types and states by scanning the genome for informative regions and providing a robust approach to quantifying methylation within these regions.

Pretraining powerful deep learning models requires large, comprehensive training datasets, which are often unavailable for medical imaging. In response, the universal biomedical pretrained (UMedPT) foundational model was developed based on multiple small and medium-sized datasets. This model reduced the amount of data required to learn new target tasks by at least 50%.

Comparative transcriptomics of whole brains across species is vital in neuroscience. Here, authors develop a deep learning method, BrainAlign, to align spatial transcriptomics across human and mouse brains. BrainAlign identifies conserved brain regions and uncovers similar patterns for marker genes.

Tiled amplicon sequencing is an essential tool for tracking the spread and evolution of pathogens, including SARS-CoV-2, however existing methods for tiled amplicon design require slow and costly downstream manual optimization. Here the authors present Olivar, a first step towards fully automated, variant-aware design of tiled amplicons for pathogen genomes.

biVI models the biophysical processes generating nascent and mature single-cell transcriptomes using variational autoencoders.

Single-cell bisulfite sequencing enables the genome-wide quantification of DNA methylation at single-cell resolution, but methods to analyze the resulting data are lacking. The MethSCAn software accurately distinguishes cell types and states by scanning the genome for informative regions and providing a robust approach to quantifying methylation within these regions.

Pretraining powerful deep learning models requires large, comprehensive training datasets, which are often unavailable for medical imaging. In response, the universal biomedical pretrained (UMedPT) foundational model was developed based on multiple small and medium-sized datasets. This model reduced the amount of data required to learn new target tasks by at least 50%.

Software is much more than just code. It is time to confront the complexity of licenses, uses, governance, infrastructure and other facets of software in science. Their influence is ubiquitous yet overlooked.

The annotation of structural glycomics data is a bottleneck in gaining insight into complex carbohydrates. An artificial intelligence model, CandyCrunch, has now been developed that accelerates the annotation process by orders of magnitude and achieves high accuracy in the prediction of glycan structures from tandem mass spectrometry data.

Fire predictions pushed to locals’ smartphones could save forests, and lives, say researchers.