Protein multiscale modelling workflows – from formulation development to protein engineering

Abstract

The activity of any given protein, whether in cell or in vitro, relies on a complex network of interactions among this protein and other molecules, such as other proteins, water, and small molecular cosolvents/excipients. These interactions occur at different spatial and temporal scales, spanning about 10 orders of magnitude in the space and 15 orders of magnitude in the time domain. As such, many different modelling techniques, each suitable for a particular spatiotemporal regime, are routinely used. However, a single process often spans more than a single time or space scale. Thus, the necessity arises for combining different modelling and simulation techniques in multiscale workflows. In this work, structurally and functionally diverse proteins such as immunoglobulins, transcription factors, autophagy receptors and non-LTR retrotransposon, were treated using several different structure-based methods (multiscale molecular dynamics simulation, molecular docking, and cosolvent based approaches) in order to study their stability and propensity to aggregate, misfold and small molecule allosteric binding. The central objective of this work was to evaluate applicability of those structure-based computational workflows in studies of protein “druggability” and development of formulations of protein-based therapeutics. First, effects of cosolvents/excipients on stability and formulation of protein therapeutics have been studied. In this part of my dissertation, multiscale (all-atom and coarse grain) simulations of monoclonal antibodies (mAbs) were carried out in different concentrations of excipients such as amino acids, salt, and sugars to check the stabilizing conditions in different formulations. Another aim of this work was to assess whether the Fab fragment may be used as a representative for the full length mAb protein in computational studies. Next part of my work was to assess the applicability of cosolvent-based workflows, such as solvent-mapping and cosolvent MD simulations in finding the stabilizing “hotspots” that contribute to allosteric regulation of estrogen receptors (ER) by small molecules. Herein, solvent mapping was used to investigate the protein protein interaction and ESR allosteric activation by methylimidazolium ionic liquids (MILs). This part of work has been carried out in collaboration liver toxicology group at Newcastle University, supporting their experimental findings on MILs acting as endocrine disruptors. Finally, I have addressed the assembly of two regulatory coiled-coil proteins, namely LINE1 ORF1p and NDP52, which are regulated by redox environment, post-translational modifications, protein-protein interactions, and cosolvents such as divalent cations. This part of work included modelling of full-length proteins and studies of protein-protein interactions in different environments. The assembly model of NDP52, which supports experimental data has been proposed. The model of Orf1p regulation, recapitulating its modulation by site-specific phosphorylation, copper and Pin1 interactions has been developed. In addition, computational site directed mutagenesis has been applied in order to validate the models.