The increasing occurrence of antibiotic-resistant bacteria and the dwindling antibiotic research and development pipeline have created a pressing global health crisis. Here, we report the discovery of a distinctive antibacterial therapy that uses visible (405 nanometers) light-activated synthetic molecular machines (MMs) to kill Gram-negative and Gram-positive bacteria, including methicillin-resistant Staphylococcus aureus , in minutes, vastly outpacing conventional antibiotics. MMs also rapidly eliminate persister cells and established bacterial biofilms. The antibacterial mode of action of MMs involves physical disruption of the membrane. In addition, by permeabilizing the membrane, MMs at sublethal doses potentiate the action of conventional antibiotics. Repeated exposure to antibacterial MMs is not accompanied by resistance development. Finally, therapeutic doses of MMs mitigate mortality associated with bacterial infection in an in vivo model of burn wound infection. Visible light–activated MMs represent an unconventional antibacterial mode of action by mechanical disruption at the molecular scale, not existent in nature and to which resistance development is unlikely.

Prediction of side chain conformations of amino acids in proteins (also termed "packing") is an important and challenging part of protein structure prediction with many interesting applications in protein design. A variety of methods for packing have been developed but more accurate ones are still needed. Machine learning (ML) methods have recently become a powerful tool for solving various problems in diverse areas of science, including structural biology. In this study, we evaluate the potential of deep neural networks (DNNs) for prediction of amino acid side chain conformations.We formulate the problem as image-to-image transformation and train a U-net style DNN to solve the problem. We show that our method outperforms other physicsbased methods by a significant margin: reconstruction RMSDs for most amino acids are about 20% smaller compared to SCWRL4 and Rosetta Packer with RMSDs for bulky hydrophobic amino acids Phe, Tyr, and Trp being up to 50% smaller.

Prediction of side chain conformations of amino acids in proteins (also termed 'packing') is an important and challenging part of protein structure prediction with many interesting applications in protein design. A variety of methods for packing have been developed but more accurate ones are still needed. Machine learning (ML) methods have recently become a powerful tool for solving various problems in diverse areas of science, including structural biology. In this work we evaluate the potential of Deep Neural Networks (DNNs) for prediction of amino acid side chain conformations. We formulate the problem as image-to-image transformation and train a U-net style DNN to solve the problem. We show that our method outperforms other physics-based methods by a significant margin: reconstruction RMSDs for most amino acids are about 20% smaller compared to SCWRL4 and Rosetta Packer with RMSDs for bulky hydrophobic amino acids Phe, Tyr and Trp being up to 50% smaller.