Phase equilibria are at the heart of many properties of substances, such as their solubility, manufacturability, and stability. They are of significant industrial and commercial interest, perhaps most importantly to the pharmaceutical industry where drug stability and solubility are two of the largest challenges of drug development. The focus of this thesis then was to develop a molecular level understanding of phase equilibria, and produce tools and models to predict phase stability. An emphasis was given to exploring solid-solid and solid-liquid equilibria and stability. Specifically, the work presented here aimed to elucidate what drives the formation of multicomponent crystals, improve available models for exploring phase equilibria phenomena and explore solubility prediction from first principles as a potentially more powerful alternative to correlation based methods. These three fundamental areas were explored by employing molecular simulation in combination with the machinery of statistical mechanics, utilising advanced sampling methods and free energy calculations. This approach has led to the development of a foundation for understanding multicomponent crystal formation in terms of molecular affinities and packing, the characterisation of a set of soft coarse-grained potentials for use in phase equilibria studies, which overcome the main limitations of the most widely used potential, and finally, the development of a novel method for solubility prediction from first principles. Here, this novel method was successfully applied to an ionic (aqueous sodium chloride) and small molecular (urea in methanol and aqueous urea) system. In the future, these results are expected to lead to a set of guidelines for predicting (and perhaps prohibiting) multicomponent crystal formation, the development of a higher class of coarse-grained transferable force field with utility in studying phase equilibria, and powerful approach for predicting solubility of even large, flexible molecules (such as pharmaceuticals).