Computational modeling of copolymer-protein interactions
Ludwig Ohl · DuEPublico (University of Duisburg-Essen) · 2019
In this computational work, I developed models that describe the interactions of proteases and random copolymers. For the development of the models, I created the framework PolyLibScan with which new models can be set up, validated and improved. The modulation of proteins with copolymers is an emerging field that has sparked much interest but as of now, there are only few computational contribution to the field of copolymer-protease interactions. Because there are still many open questions, it is important to develop new tools that can help the understanding in this field. The high conformational flexibility of the copolymers and also their large variability of their random sequences pose a challenge to successfully explore conformational- and sequence-space. Because of this challenge, the model relies on coarse grained simulations to reduce complexity and make computation feasible. For the calibration and validation of the models, I use the results of the inhibition experiments of Gilles et al., who found numerous inhibitors among 49 synthesized random copolymer-types for six proteases. For the coarse graining, the most minimalistic approach is chosen as a starting point by using a single united atom per monomer and amino acid. The inhibition property of each copolymer-type is determined by multiple simulations. The first model relies on two computational methods to identify binding sites for the monomers and the proteases, which are used to parameterize the model. Because this model is not accurate enough, a second model is developed that relies on the charges of monomers and amino acids for the interactions. This model has a better agreement with the experiments for three of the proteases. Based on these results, the third model adds the hydrophobic interaction, more precisely takes into account the implicit solvent and makes the protonation of the particles pH-dependent. With these changes, the third model is a further improvement over the second model. The last model changes the coarse grained structure of the copolymers to more closely resemble the all-atom copolymers. Even though the results are different to the last model, it is similar in its accuracy to the third model. Despite their minimalistic structure, the last two models accomplish a good agreement with four of the six protease, but also show the limits of single bead models with the other cases.