Cell factories converting bio-based precursors to chemicals present an attractive avenue to a sustainable economy, yet screening of genetically diverse strain libraries to identify the best-performing whole-cell biocatalysts is a low-throughput endeavor. For this reason, transcriptional biosensors attract attention as they allow the screening of vast libraries when used in combination with fluorescence-activated cell sorting (FACS). However, broad ligand specificity of transcriptional regulators (TRs) often prohibits the development of such ultra-high-throughput screens. Here, we solve the structure of the TR LysG of Corynebacterium glutamicum, which detects all three basic amino acids. Based on this information, we follow a semi-rational engineering approach using a FACS-based screening/counterscreening strategy to generate an l-lysine insensitive LysG-based biosensor. This biosensor can be used to isolate l-histidine-producing strains by FACS, showing that TR engineering towards a more focused ligand spectrum can expand the scope of application of such metabolite sensors.

Molecular docking and molecular dynamics (MD) are powerful tools used to investigate protein-ligand interactions. Molecular docking programs predict the binding pose and affinity of a protein-ligand complex, while MD can be used to incorporate flexibility into docking calculations and gain further information on the kinetics and stability of the protein-ligand bond. This review covers state-of-the-art methods of using molecular docking and MD to explore protein-ligand interactions, with emphasis on application to drug discovery. We also call for further research on combining common molecular docking and MD methods.

Computational enzyme design has made great strides over the last five years. Traditional methods of enzyme design require synthesis and evaluation of many mutations. Computational enzyme design has emerged as a powerful tool to predict how specific mutations modify a protein’s activity, stability, and/or selectivity. Such computational approaches can evaluate many mutations and reduce the load of in vitro work by identifying mutations likely to accomplish design objectives. Computational approaches can explore mutational spaces inaccessible in traditional mutagenesis. Computational methods reduce cost and time compared with experimental approaches. We review the efficacy and key differences of computational enzyme design methods as published in recent studies. The included articles used computational methods to design enzymes, were published no earlier than 2015, met design objectives, and verified results in vitro.

Molecular docking tools are regularly used to computationally identify new molecules in virtual screening for drug discovery. However, docking tools suffer from inaccurate scoring functions with widely varying performance on different proteins. To enable more accurate ranking of active over inactive ligands in virtual screening, we created a machine learning consensus docking tool, MILCDock, that uses predictions from five traditional molecular docking tools to predict the probability a ligand binds to a protein. MILCDock was trained and tested on data from both the DUD-E and LIT-PCBA docking datasets and shows improved performance over traditional molecular docking tools and other consensus docking methods on the DUD-E dataset. LIT-PCBA targets proved to be difficult for all methods tested. We also find that DUD-E data, although biased, can be effective in training machine learning tools if care is taken to avoid DUD-E's biases during training.

The prediction of amino acid contacts from protein sequence is an important problem, as protein contacts are a vital step towards the prediction of folded protein structures. We propose that a powerful concept from deep learning, called ensembling, can increase the accuracy of protein contact predictions by combining the outputs of different neural network models. We show that ensembling the predictions made by different groups at the recent Critical Assessment of Protein Structure Prediction (CASP13) outperforms all individual groups. Further, we show that contacts derived from the distance predictions of three additional deep neural networks—AlphaFold, trRosetta, and ProSPr—can be substantially improved by ensembling all three networks. We also show that ensembling these recent deep neural networks with the best CASP13 group creates a superior contact prediction tool. Finally, we demonstrate that two ensembled networks can successfully differentiate between the folds of two highly homologous sequences. In order to build further on these findings, we propose the creation of a better protein contact benchmark set and additional open-source contact prediction methods.