| Peripheral membrane proteins are crucial for many cellular processes such as cellular signaling and membrane-trafficking. They bind various membranes reversibly in response to a certain signal and it is these proteins that are the focus of this thesis, Broadly, the goal of the research presented here is a comprehensive sequential, structural and dynamical analysis of these proteins through computational took and bioinformatics approaches. First, a machine learning protocol is developed using the key structural features of known peripheral proteins to predict the binding properties of unknown ones. As a first step towards their genome-scale identification, the analysis is then extended to sequence-based features using a special class of unsupervised machine learning technique developed. The determinants behind the binding of these proteins are then illuminated using a set of simple rules that can be easily interpreted in biological context. Following that, computational techniques such as molecular dynamics (MD) simulation are used to get a better comprehension of the mechanism of their interaction with the membrane such as elucidation of favorable conditions, important cofactors etc. With three case-studies of peripheral domains, MD is used to elicit the relative importance of various co-factors in their binding, the mechanism of compensation of one co-factor by another and the mechanism of the unbinding from the membrane, Other computational tools such as docking and electrostatic calculations are also used to elucidate other details about them such as their membrane binding orientation. Finally, an exclusive and publicly-available resource dedicated to the peripheral domains and host proteins is constructed providing all the information about them. |
