Drug screening articles from across Nature Portfolio

An explainable deep learning model enables the identification and virtual optimization of antimicrobial peptides (AMPs) from the oral microbiome, yielding novel AMPs with activity against ESKAPE pathogens and demonstrating efficacy in a mouse wound infection model.

Shedding light onto the role of phosphorylation in TDP-43 cytoplasmic mis-localization.

Mapping the metabolic effects of drugs helps define their mode of action.

An explainable deep learning model enables the identification and virtual optimization of antimicrobial peptides (AMPs) from the oral microbiome, yielding novel AMPs with activity against ESKAPE pathogens and demonstrating efficacy in a mouse wound infection model.

Sample, financial and labor requirements are key barriers to scaling up high-content phenotypic discovery efforts. A broadly applicable method overcomes these challenges through experimental compression (by pooling various perturbations) and computational deconvolution (of their individual effects), empowering the use of phenotypic screening to advance therapeutic discovery.

In this Tools of the Trade article, Xinwen Liu describes the development of VIBRANT, a vibrational spectroscopy method for high-content phenotypic profiling, and highlights its use to predict drug mechanisms of action or identify potential drug candidates.

Revealing a drug’s mechanism of action (MOA) is costly and time-consuming. In this study, we used deep learning to extract temporal mitochondrial phenotypic features after exposure to drugs with known MOAs using re-identification algorithms. The trained model could then predict the MOAs of unidentified substances, facilitating phenotypic screening-based drug discovery and repurposing.

Aficamten, a small-molecule myosin inhibitor, binds to the catalytic domain of cardiac myosin and reduces contraction by stabilizing the initial weak interaction between myosin and actin.

Improving the precision of gene correction through RNA base editing typically involves a trade-off with efficiency. We find using G•U wobble base pairs is a universally applicable strategy to improve editing precision and efficiency, and is effective in a mouse model of Rett syndrome in vivo.