Investigating the molecular origin of the hydrophobic effect with a model of molecular connectivity

Molecular technologies offer the prospect of creating new materials and new organisms by the process of molecular manipulation. For commercial viability we must harness the power of molecular self-assembly which, in turn, demands that we must first understand it. Molecular systems are complex systems that can be understood as networks of interactions between molecules. However, in molecular systems all molecules can potentially interact in different ways simultaneously. The challenge is then to understand the process of molecular self-assembly by determining which aspects of these interactions are important to the outcome. Arguably, the major contributor to molecular self-assembly in aqueous systems is the spontaneous aggregation of lipids (non-polar molecules), commonly termed the hydrophobic effect. The hydrophobic effect is responsible for many molecular self-assembly phenomena including membrane assembly, protein folding and bubble or foam formation. Despite decades of investigation, the molecular origins of this effect are still being debated. It is known that water molecules drive the hydrophobic effect, and that the tetrahedral arrangement, of the fluctuating hydrogen-bond network formed by water molecules, is responsible for many of liquid water’s anomalous properties. Debate continues whether the hydrophobic effect is a result of the hydrogen bonds formed between water molecules (connectivity theory) or simply a result of the relatively small size of water molecules (small-size theory). What appears clear is that the properties of the hydrogen bond network could be important not only to liquid water, but also to understanding the hydrophobic effect. Additionally, if the network properties are important, as network theory suggests, connectivity preference could be important. Molecular dynamics is a deterministic computer simulation technique commonly used to investigate chemical system dynamics. This modelling technique assumes a “uniformist” approach, where all molecules interact via the same potential energy function, and where coordination constraints are imposed implicitly. This approach, precludes modelling a system in terms of a network, where the rules of connectivity, and maximum coordination, are explicit properties of the model. Therefore, the first step in determining whether network properties are important to the liquid water, and the hydrophobic effect, is to investigate a new chemical modelling technique that can explicitly model a network. This thesis proposes the Fluctuating Network algorithm (FN) as an extension of the well-known 3D Molecular Dynamics Algorithm (MD) as a means of explicitly modelling a chemical network. The FN algorithm focuses on the interactions and the aspects of each interaction important to network connectivity. It therefore distinguishes bonded (strong hydrogen bond) from non-bonded (weak hydrogen bond) interactions, explicitly enforces maximum coordination for bonded interactions and, detects conditions under which interactions can viably change from bonded to non-bonded and vice versa. This is achieved by extending the MD algorithm to allow hydrogen-bond interactions to have two states - bonded and non-bonded - with each state employing a distinct potential energy function. Exchange between bonded and non-bonded states is facilitated through a network reorganisation procedure which is performed each time-step after molecule positions are updated. This procedure allows the network connections to reflect changing molecular positions and maximise energetically favourable interactions. A model employing the FN algorithm was then developed and verified to model water molecules. Simulations of FN-water, with simple potential functions, successfully reproduced many of the properties of liquid water and compared favourably to more complex models like TIP4P. FN-water/lipid mixtures all reproduced the hydrophobic effect; however, it was found that Lorentz rule deviations (excluded-volume effects) were a better predictor of solute hydration or aggregation than solute size. It was found that excluded-volume effects dominate liquid water; however, the structuring produced by hydrogen bonding would allow connectivity preference to play a significant role under certain conditions. This thesis is inherently multidisciplinary. It makes contributions to both chemistry and computer science. Its contributions to chemistry are as follows: 1. This thesis proposes the fluctuating network algorithm (FN Algorithm). This algorithm demonstrates how increased algorithmic complexity can facilitate simulation of complex interactions in a molecular context (like charge-transfer interactions), while reducing the complexity of the potential functions used to approximate the interactions and therefore potentially decreasing computational expense for molecular simulations with these interactions. 2. This thesis demonstrates that a network model of water can reproduce many of the structural and dynamic properties of liquid water. 3. This thesis demonstrates that this network model of water combined with hard-sphere solute models can reproduce the hydrophobic effect. 4. The experimental approach taken by this thesis reveals that in dilute solute concentrations solute aggregation and hydration is highly sensitive to deviations from the Lorentz rule (excluded volume effects), and only slightly sensitive to solute size. As solute concentration increases, the sensitivity to solute size increases, whilst the sensitivity to deviations from the Lorentz rule decreased. 5. This thesis proposes a preliminary theory on the mechanism of the hydrophobic effect - distinct from the small-size theory and the connectivity theory - which takes into account the Lorentz rule findings. This theory explains the solubility of small and large hydrophobic molecules, and the effect of hydrogen bonding groups, whilst addressing why deviations from the Lorentz rule drives solute aggregation and hydration. The computer science contributions are as follows: 1. This thesis demonstrates the potential of connection preference rules, in 3D dynamic network environments, to drive mixing or de-mixing. 2. This thesis demonstrates that self-assembling systems, which rely on weak interaction forces, have a critical density threshold for simulation. This thesis demonstrates successful application of computer science concepts to solve interdisciplinary problems in chemistry.