Development of novel computational methods suitable for modelling intrinsically disordered proteins

Abstract

Proteins without a stable tertiary structure are known as intrinsically disordered or metamorphic. These proteins denoted as IDPs – or protein domains denoted as IDRs – exert crucial roles in cellular signalling, growth and molecular recognition events. Due to their high plasticity, IDPs and IDRs are very challenging for experimental and computational structural studies. To enable these, all-atom molecular dynamics (MD) simulations are used, as they provide insight into structure and dynamics at the atomistic level of detail. However, the current generalist physical models (protein force fields and solvent models) used in MD simulations are unable to generate satisfactory ensembles for IDPs/IDRs when compared to existing experimental data. This work aimed to improve on the state-of-the-art accuracy for simulations of IDPs/IDRs without sacrificing accuracy for folded domains. Herein, the accuracy of several different force fields frequently used for simulations of proteins was compared, in simulations of both ordered and disordered systems. The results showed that each force field has strengths and limitations. Given the fact that interactions with the solvent are pivotal for accurate simulations of intrinsically disordered proteins, a novel solvation model was developed, denoted as Charge-Augmented 3 Point water model for Intrinsically disordered Proteins (CAIPi3P). CAIPi3P model was generated through systematic scanning of the dipole moment values calculated for the popular TIP3P three-point water model. By increasing the dipole magnitude, the agreement between experimental and calculated small-angle X-ray scattering (SAXS) curves was massively improved for a series of model IDRs. To further improve the simulations of proteins containing IDRs, a novel method to assemble force field parameters has been developed. Denoted as Hybrid_FF, it merges parameters from different established force-fields, performing well for structured and disordered regions (AMBER99SB-ILDN and AMBER03ws, respectively), parametrising each secondary structure differently. Testing these joint parameters for a series of IDR-containing proteins showed that such an approach improved the accuracy of the sampled configurations for long disordered regions. Finally, a software to estimate and analyse the transition dynamics of intrinsically disordered regions has been developed in this work. Named structural quantifier of entropy (SQuE), it uses a first-order approximation to the probability distribution to assess the structural entropy for protein transitions barriers. It is expected that tools developed in this study will generate more accurate IDP/IDR ensembles, broadening the range of biologically relevant systems amenable to atomistic molecular dynamics simulations.