Protein design originally began as a field intended to demonstrate our understanding of how proteins fold. Over the last three decades, advances in design principles and algorithms have paved the way for de novo computational design of new classes of proteins. While biological or drug-like applications for these proteins have been demonstrated, existing computational techniques are generally not able to predict whether a design will be highly or only moderately functional. We have used computational simulation and modelling to study the mini-fluorescence activating-protein (mFAP) that was designed to function as a fluorescent marker or biosensor when bound to an exogenous chromophore, 3,5-difluoro-4-hydroxybenzylidene imidazolinone (DFHBI). DFHBI fluoresces when it is in a planar-Z conformation and mFAPs stabilize DFHBI in this configuration, increasing fluorescence 10-100x fold relative to free DFHBI in solution. However, when compared to eGFP they are about 5-fold less efficient (i.e., dimmer). In our latest work, we have built a predictive model that accurately captures the photophysical behavior DFHBI in the binding pocket of 13 mFAP designs via molecular dynamics (MD) simulations. When analyzing the brightest mFAP variant, we have found that certain side-chain rotational isomers (rotamers) are strongly associated with keeping DFHBI in a planar-Z conformation and thus a preference for conformationally shifting towards functional states. Because of the high computational cost for running MD simulations, we are in the process of developing a high-throughput, low-cost algorithm—using the recently developed MDR-FEP method—to find mutants that increase the existence of these rotamers.
Read full abstract