Data mining articles from across Nature Portfolio

Single-cell multi-omics integration across gene expression, protein abundance and chromatin accessibility remains challenging. The authors present scMAGCA, an interpretable framework for integrating these modalities, and apply it to Alzheimer’s disease and kidney cancer to identify disease-associated cell states and biomarkers.

scMethCraft is a computational toolkit for analysing single-cell DNA methylation data. It supports tasks such as cell clustering, data integration and data enhancement, enabling the discovery of epigenetic patterns across cell types.

Single-cell multimodal data integration presents a substantial challenge due to the heterogeneous and incomplete modality compositions. Here, the authors introduce Palette, a flexible and interpretable data integration method, enabling the efficient alignment of complex multimodal datasets.

Here the authors develop a machine-learning tool (MALCA) that uses inhibition zone diameters from routine disc diffusion tests to directly detect carbapenemase-producing Enterobacterales and identify the carbapenemase type with high accuracy, enabling earlier targeted treatment without additional laboratory resources.

Traditional single-cell models often overlook the proteomic context of cell phenotypes. Here, authors present CAPTAIN, a multimodal foundation model pretrained on 4.2 million cells. It aligns RNA and protein profiles to enhance cell annotation, protein imputation, and intercellular discovery.

Despite the proven therapeutic efficacy, CAR T cell treatment carries the risk of severe toxicities. How sex-associated differences affect therapy-associated toxicities is underexplored. Here, the authors conduct a retrospective pharmacovigilant study using the FAERS database and highlight sex-associated differences in CAR T therapy complications, with females exhibiting a consistent pattern of increased adverse events across cancer types and therapy products.

A comprehensive assessment of how transformation of dietary nutrients by the gut microbiota influences host health is needed. Systematic integration of 5,554 global gut metagenomes and various biochemical databases reveals how dietary phytonutrients are biotransformed by the gut microbiota and their enzymes, and how these biotransformations relate to human health.

Mass spectrometry data analysis has long been limited to known molecules and exact matches. In a recent manuscript, a scalable search algorithm is proposed for uncovering both known compounds and novel molecular variants, enabling insights into natural product biosynthesis.

The bankruptcy of 23andMe was an inflection point for the direct-to-consumer genetics market. Although the privacy of consumer data has been highlighted by many as a concern, we discuss another key tension in this case: the corporate enclosure of scientific data that has considerable potential value for biomedical research and public health.

A deep-learning algorithm unravels a collection of archaeasins, peptides from the archaeal proteome with potential antimicrobial activity and implications for the development of next-generation antibiotics