Prediction of multiscale thermodynamic phenomena in chemical formulations screening via a novel molecular modeling framework
Abstract
The design of stable chemical formulations is hindered by the entangled and nonlinear nature of molecular interactions as well as by the practical limitations of experimental exploration and characterization. Theoretical and numerical modeling provides the means to unravel this complexity. Specifically, modeling enables the decoupled study of interactions (i.e. studying each molecular interaction independently), revealing mechanisms, and systematically exploring the immense chemical space to design and develop formulations that meet targeted properties. To achieve that, in this work we have developed a multiscale molecular modeling framework designed to understand, predict and explore the chemical space for solvation, aggregation and interfacial properties. The developed framework establishes a direct link between quantum chemistry and mesoscale physics, unifying these traditionally separate domains into a continuous framework. To create such a framework, we had to first build a solution that connects quantum chemistry with statistical thermodynamics to predict molecular interactions in a fast and accurate fashion, something crucial in our efforts to explore a big part of the chemical space. Subsequently, we had to modify the Dissipative Particle Dynamics mesoscale solver making it capable of studying protein-protein interactions and interfacial properties.
Our first methodological advancement concerns the development of a semi-empirical quantum chemical protocol capable of accurately predicting molecular electrostatic profiles and free energy of mixtures. This approach combines a bond charge correction introduced in the AM1 semi-empirical Hamiltonian, with the COSMO-RS solvation model and introduces an automated internal hydrogen bonding (IHB) sampling algorithm. This combination drastically accelerates the computation of molecular electrostatic profiles without sacrificing the fidelity of Density Functional Theory (DFT). The inclusion of internal hydrogen bonding proved to be crucial, as neglecting them systematically underestimates molecular partitioning and permeability. The protocol demonstrated that molecules capable of intramolecular hydrogen bonding exist as ensembles of various conformations whose populations depend on the polarity of the solvent. Proper statistical inclusion of these conformations is essential for reaching quantitative accuracy. In addition, the work underscored the inherent complexity of capturing specific interactions for certain functional groups, indicating that further investigation is needed to identify which additional descriptors are required for their accurate representation in predictive models.
The second methodological advancement extends predictive modelling from small molecules to macromolecular and colloidal systems through a modified Dissipative Particle Dynamics (DPD) model that introduces a physically meaningful attractive conservative potential. This extension allows the DPD formalism, which is typically limited to purely repulsive interactions resulting in super-condensed gas representations, to model systems in STP conditions, capturing phase transitions, interfacial phenomena and protein-protein interaction interfaces. Our protocol successfully predicts the titration behaviour of proteins using Poisson-Boltzmann electrostatics and reproduces the effect of ionic strength on protein-protein interactions and colloidal stability. By coupling these electrostatic calculations with mesoscale DPD simulations, the framework identifies the balance between electrostatic repulsion, hydration, and cohesive forces that governs aggregation. Therefore, it can guide experimentation towards which buffer, pH, and ionic strength should be selected to ensure a soluble solution. This approach makes it, also, possible to map the protein-protein aggregation interface, providing a foundation for optimizing formulation design: additives can be introduced in silico at this interface to inhibit aggregation and enhance solubility. These capabilities transform mesoscale simulations from descriptive to predictive tools in the design of biological and chemical formulations. The newly introduced conservative DPD potential enables the study of interfacial systems, allowing the direct simulation of wetting and de-wetting processes on polymeric substrates and fluorinated coatings. The new model supports the study of phase coexistence and surface tension, phenomena that are absent in conventional DPD. Thus, it permits the calculation of static and dynamic contact angles. Simulations reveal that for coatings to reach optimal repellence, they require not only chemical hydrophobicity but also structural rigidity. Flexible hydrophobic chains, though low in surface energy, dynamically reorient under droplet motion, increasing contact angle hysteresis. Rigid hydrophobic conformations maintain stable orientation, resulting in lower hysteresis and improved repellent efficiency. The dynamic surface energies obtained from simulation reveal the chemical and structural heterogeneity of the coatings, showing that molecular reorientation is the key microscopic mechanism underlying macroscopic wetting behaviour. In this way, the framework connects molecular structure with interfacial thermodynamics. The results of this thesis demonstrate that predictive formulation design requires the deep understanding of molecular interactions. The presented framework achieves this by transferring electronic-structure accuracy to mesoscopic simulations, allowing one to predict, interpret, and optimize thermodynamic behaviour across a vast chemical composition space. It shows that once molecular mechanisms can be decoupled and quantified individually, their thermodynamic consequences can be systematically tuned, enabling informed optimization of multicomponent systems. Beyond the methodological advances, the framework provides scientific insights into phenomena ranging from solvation and permeability to protein aggregation and interfacial wetting, offering a physicochemical understanding of formulation stability and performance.
