Assessing the Adsorption of Various Proteins with Conformational Screening Utilizing k-Means Clustering and CHARMM Potential Energy Calculations
Date of Award
Doctor of Philosophy (PhD)
Utilizing Quartz Crystal Microbalances (QCMB) to study the interaction of biochemical systems at interfaces is a diverse field with many potential applications. However, with an unfunctionalized surface, this technique suffers from a lack of known binding specificity which can denature the protein and result in a loss in activity. Under such conditions, investigating the adsorption mechanism of proteins to hydrophobic surfaces is difficult with traditional molecular dynamics simulations. In this work, we propose a screening model using geometric transformations, k-means clustering, and MD simulations to investigate the clustered orientations. In order to validate this model and predict the surface density of Concanavalin A adsorbed to graphene, we will attempt to reproduce the empirical QCMB mass of 671$\pm$21ng/cm$^2$ in the work of Alva et. al. The first phase of the model involves a series of geometric transformations combined with CHARMM potential energy calculations to sample the potential energy surface under the Generalized Born Implicit Solvent. Protein topology and potential energy information from the geometric transformations was input into a k-means clustering algorithm to identify a set of clusters representative of the conformational space.
Molecular dynamics simulations conducted from the representative k-means cluster set displayed varying degrees of surface density with one conformer, nearest the cluster centroid, consistent with an empirical adsorbed surface density of $676.915\pm8.250$ ng/cm$^2$. The profile of adsorption for the low energy conformers would suggest a mechanism of adsorption that leverages hydrophobic carbohydrate binding residues within the protein to stabilize adsorption to graphene. Results suggest the conformational screening method can provide insight into the mechanisms of adsorption for an unfunctionalized surface and target protein.\\
In a related study we aim to use this sampling method to model the adsorption of the human insulin dimer and monomer to graphene. Using the surface area of the protein near graphene and CHARMM potential energy of each orientation k-means clustering was used to determine the similarity of different orientations. Clustering of the insulin dimer indicates that 6 orientations are minimally representative of the overall conformational space. Using the orientation nearest the cluster centroid, as an average representation of that cluster, molecular dynamics were run on the cluster set. Once complete, the initial set of MD simulations were re-sampled by following geometric trends in the relative binding energies. The results of this re-sampling show one conformation 111:306 with the lowest relative binding energy of -75.96 $\pm$ 16.82kcal/mol and a surface density of 351.05 $\pm$ 5.77ng/cm$^2$. Topology of the binding site shows association with aromatic and hydrophobic residues TYR14, PHE73, and ILE10 which are expected to stabilize the adsorption of this conformation. Results from the insulin monomer indicate four orientations are representative of the attribute space one of which (39:188) displays the lowest relative binding energy of -61.700$\pm$7.893 kcal/mol. This conformer displayed an affinity toward the classical binding surface for the insulin receptor and a predicted surface density of 18.955$\pm$0.256 ng/cm$^2$. \\
Monolayer graphene, aside from its hydrophobicity and adsorption effects, has unique physical properties which make it a good candidate for a membrane material. Despite its one atom thickness, graphene has a high mechanical rigidity with a Young's Modulus of 1.13 TPa. Using electron beams or highly charged ions, one can create graphene membranes with varying sizes and pore functional groups. In this work we investigate the application of graphite as pressure gated membranes generated via low angle ion impacts. We hypothesize that, with the correct structure, these shutter type membranes can act as filters, opening under high pressure gradients and closing via the self-retraction mechanism of graphite. Initial results indicate the mechanical properties of graphene are modeled to an empirical accuracy within the CHARMM potential. However, the Lennard-Jones potential fails to accurately represent the intra-layer potential of graphite. Additional membrane structures will need to be investigated to prevent native opening of the graphite pore. One proposed membrane structure would include pores in the graphite membrane less than 0.5nm in diameter to prevent water transport while providing an energetic penalty for opening the shutter pore. Based on current results further work should be conducted with the Kolmogorov potential and pore properties investigated via Density Functional Theory.
Overstreet, Richard Edwin, "Assessing the Adsorption of Various Proteins with Conformational Screening Utilizing k-Means Clustering and CHARMM Potential Energy Calculations" (2019). All Dissertations. 2486.