Reviews & Analysis

CRISPR base- and prime-editing pooled screens reveal the function of genetic variants at unprecedented resolution.

Anderson and colleagues discuss the progress and challenges of using mRNA for vaccines and immunotherapy.

Artificial intelligence finds candidate peptide antibiotics in the human gut microbiome.

The jellyfish-derived green fluorescent protein StayGold is bright and hardly fades, contributing to improving spatiotemporal resolution and dramatically extending the observation period. To fully benefit from the rich photon budget, we tried some unusual illumination modalities for sustainable, quantitative live-cell or volumetric imaging.

We developed EasyFuse, a computational machine learning pipeline that detects cancer-specific gene fusions with superior performance over existing tools. Individual gene fusions exhibit a high frequency of pre-established CD4⁺ and CD8⁺ T-cell responses and thus represent a previously untapped source of neo-antigens that can be exploited for personalized immunotherapies.

A new assay for cell-free DNA fragmentation improves prediction of gene expression in liquid biopsies.

A thorough study of lab-made pancreatic beta cells reveals their differences from primary cells.

Using a combination of metagenomic big data and deep learning tools, small proteins that inhibit pathogens — and could be further developed into novel antibiotics — are mined en masse. Such methods could greatly improve the throughput of drug discovery and translational usage of the microbiome.

A new method uses synthetic introns to express therapeutic proteins selectively in cells bearing cancer-initiating mutations affecting RNA splicing factors, while healthy cells remain unaffected. This approach enabled the eradication of human leukemia, breast cancer and uveal melanoma cells in mouse models and significantly prolonged host survival.

Producing commodity chemicals in bacteria that live on industrial air pollution captures more greenhouse gases than it emits.

Regev, Theis and colleagues outline the challenges and concepts of the analysis of spatial transcriptomics data.

A deep learning algorithm for protein structure prediction is used in reverse for de novo protein design.

Tagging peptides with DNA allows repeated reads via a helicase in a nanopore with reduced error rates.