WG1: Computational optimization of global enzyme properties

Stability engineering of proteins has traditionally focused on small reversibly folding model proteins. The larger enzymes introduce additional challenges, which can include finding the critical regions for stability, dependence of stability on cofactor binding, or limitations to what mutations can be introduced to stabilize the protein without inhibiting its function by rigidification of the active site. Some of these challenges are being addressed by algorithms that are broadly known as machine learning. These methods do require large data sets of mutations for which the effect of stability has been characterized; but more and more of such homogeneous sets of characterized mutations are becoming available and also some of the experimental data can often be replaced by in-silico generated data. For enzymes that require cofactor binding for their operational stability, novel protocols will require optimization of the interactions between protein and ligand, which encompasses chemically diverse interactions and are therefore more challenging to improve. The cofactors also tend to be large, which requires better conformational sampling than for most ligands. Another key property sometimes related to protein stability is protein solubility. There are currently no reliable tools for prediction or even engineering protein solubility, practically due to high complexity of phenomenon as well as lack of high-quality data. Target enzymes include (cofactor dependent) enzymes such as C-H bond activating P450s monooxygenases, peroxidases, peroxygenases, flavin-containing oxidases, amine forming enzymes, epoxide hydrolases, peptide bond forming enzymes, haloalkane dehalogenases, decarboxylases and luciferases and sugar-acting enzymes.