Investigation of solvent-mediated molecular interactions driving self-assembly

Self-assembly is the most promising technique to create atomistically accurate nanostructures in today's world. Currently, practicing self-assembly in industrial production of ordered structures is limited by poor understanding of the fundamental driving forces that dictate specific structures, underlying thermodynamics, and characteristics of specific electrostatic and dispersion forces. This dissertation focuses on investigating molecular level interactions that drive the self-assembly of simple structures such as proto-micellar aggregates of amphiphilic alcohols, or more complex case of the short polypeptides which form partially helix structures in water. The approach is to use combination of molecular dynamic simulations and statistical mechanics to examine experimentally unapproachable interactions among molecular moieties and their implications on the assembled structures.;The first problem addresses alcohol induced aggregation of the hydrophobic species in the nuclear pore complex. The nuclear pore complex (NPC) provides a size-selective transport barrier that regulates traffic between the nucleus and cytoplasm. Considerable mass of the NPC is composed of intrinsically disordered nucleoporin proteins rich in phenylalanine (F) -glycine (G) repeat units, alternatively called as the FG-Nups. The meshwork created by the disordered FG-Nups controls transport by maintaining a permeability barrier on the principles of hydrophobic exclusion. Experimentally it has been observed that alcohols of increasing hydrophobicity reversibly diminish the permeability barrier, suggesting the hydrophobic side chains of the FG-Nups microphase separate to form a mesh-like gel that sieves large species. Using molecular simulations, we investigate the solvent-mediated interactions between hydrophobic FG-Nup side chains and how they are altered after addition of alcohols. The cooperative aggregation of hydrophobic side chains is illustrated by interfacially active cosolvent alcohols such as 1,2-hexanediol. The aggregating alcohols in simulation systems are found to be similar to those disrupting NPC barrier in experiments. Simulations results suggest that the decreased selectivity of NPC is a result of holes opening within the FG-Nup mesh, providing pathways for trans-nuclear membrane transport.;In second problem, we measure interactions of hydrophobic solutes with water using solute-water pair correlations functions and their Kirkwood-Buff integrals. The Kirkwood-Buff (KB) theory of solution provides relationship between molecular pair correlations and the bulk level thermodynamic properties e.g. partial molar volume (PMV). The long-ranged and oscillatory KB integrals of the simulation pair correlation functions are suppressed with the methods of solvent smearing to calculate PMV. Solute PMV is divided into proximal volume sub-domains on each functional group and proximity approximations are used to evaluate KB integrals. Simulation partial molar volumes are systematically greater than experiment, with the difference between simulation and experiment increasing with solute size. These differences are attributed to possible deficiencies in the volumetric properties of the interaction potentials employed. It is found that the structure of water near identical groups on different solutes is effectively same, as quantified by proximal correlation functions. As a result, we propose a new group contribution correlation for the PMV rooted in Kirkwood-Buff theory and the proximity approximation.;Depending on the amino acid sequences, certain polypeptides acquire partially helical structures. The alpha-helix content in these structures is encoded in their amino acid sequence where each amino acid carries specific propensity to participate in alpha-helix structures. Typically, Lifson-Roig theory of helix-coil transition provides these propensities when fitted to fractional helicities from either simulations or circular dichroism experiments. Circular dichroism however can predict only the average alpha-helical structure and looses details of distribution of helix along the sequence. LR theory on the other hand predicts the alpha-helix probabilities at individual amino acid positions for different helical events. From simulations, accurate probabilities at each location on sequence can be determined which semiquantitatively agree with LR predictions. LR propensities also predict the thermodynamics of helix-coil transition as well as distinguish meaningfully between helical propensities of different amino acids depending on hydrophobic, hydrophilic or charged side chain.