A protein switch alters its behavior in response to a change in environmental conditions, such as the presence of a specific ligand. The ability to create such switches through engineering or modification of existing proteins has a variety of transcending biotechnological applications. Through directed evolution techniques, recent experiments have shown that domain insertion can be used to combine maltodextrin-binding protein (MBP) and TEM-1 β-lactamase (BLA) to produce a switch that activates enzymatic activity of β-lactamase in the presence of maltose. An effective structure prediction algorithm can potentially complement directed evolution techniques with structure-based design and provide insight into the mechanism of the switch. We are developing a new algorithm to predict the structure of this engineered protein by modeling the two proteins as separate host (MBP) and insertion (BLA) domains and simultaneously optimizing the insertion domain orientation and the conformation of the connecting linkers. The algorithm uses a combination of peptide fragment insertion, phi/psi perturbations, and customized Monte Carlo moves to orient the insert domain and create the connecting linkers. Monte Carlo minimization efficiently allows the backbone to relax while rotamer packing finds the optimal side chain positions. To validate the efficacy of the algorithm, we scanned a database of domain insertion proteins to create a test set of targets. We used this test set to show that the algorithm is able to start from random perturbations of the insertion domain and find the native conformation of the two domains. This data shows that the algorithm is capable of predicting the structure of proteins with naturally occurring domain insertions and indicates its potential for structure prediction of engineered proteins by domain insertion. Ultimately the program will predict the insertion domain structure from structure of separate source domains or their homologues.

Extended Abstract Status: Not Uploaded