Research Project
10275155
Project Summary The overall aim of the research is Shukla group is to develop computational methods that facilitate investigation of rare conformational transitions in proteins and help guide the design of experiments to validate the in silico predictions. In par- ticular, we apply these computational methods to investigate functional regulation of membrane proteins such as membrane transporters and G-protein coupled receptors (GPCRs). Here, we propose development of transfer learning based methods to predict the effect of mutations on protein function and apply these methods to investigate monoamine transporters, sugar transporters and Class C GPCRs. Deep mutagenesis, whereby tens of thousands of mutational effects are determined by combining in vitro selections of sequence variants with Illumina sequencing, is an emerging technology for indirectly interrogating and observing protein conformations in living cells; the solving of an integrative structure of a neuronal class C G protein-coupled receptor in an active conformation by deep mutagenesis-guided modeling is one prominent example of this approach's success. Using deep mutagenesis and molecular dynamics simulations to inform each other, we plan to determine the mechanism of ion- coupled neurotransmitter import by monoamine transporters at atomic resolution. Fluorescent substrates have enabled us to use ?uorescence-based sorting of libraries of transporter mutants to ?nd mutations along the entire permeation pathway that increase or decrease substrate import. These comprehensive mutational landscapes will be used to interpret and support/reject hypotheses from simulations, including the role of ion-coupling in substrate transport regulation, proposed free energy barriers in the conformational-free energy landscape that limit import kinetics, and how sodium-neurotransmitter symport is coupled by a shared cytosolic exit pathway. Other notable features that arise from the deep mutational scans (e.g. putative regulatory sites) will be further explored, and a machine learning algorithm will be applied to transfer mutagenesis information to related transporters; the predicted mutational landscapes will then be validated by a small number of informative targeted mutants. We will further relate sequence to conformation and activity in metabotropic neurotransmitter receptors and sugar transporters. Finally, we plan to improve the proposed transfer algorithms by using deep learning techniques, which will facilitate integration of features derived from simulation datasets and multiple deep mutational scans to inform the effect of mutations on related proteins or tasks. The success of the proposed research program of results will be measured by development of algorithms that can accurately predict the variant effects on protein structure and function, elucidation of the mechanisms of ion-coupled regulation of neurotransmitter transport, selectivity mechanisms in sugar transporters and activation mechanisms of class C GPCRs.
