Central to the function of proteins is the concept of molecular recognition. Protein–ligand and protein–protein interactions make up the bulk of the chemical processes that give rise to living things. Realizing the full potential of protein design technology will therefore require an increased understanding of the design principles of molecular recognition. We have tackled problems involving molecular recognition by using computational methods to design novel protein-ligand and protein-protein interactions. Firstly, we set out to design a protein capable of recognizing lanthanide metal ions. Protein-lanthanide systems are of interest for their potential to serve as purification agents for use under biological conditions. We have designed a highly dense 6-coordinate lanthanide binding at the core of a de novo protein, and used the dynamical aspects of the protein to achieve a degree of differentiation between elements in the lanthanide series. Secondly, we investigated systems of homo-oligomeric protein complexes that self-assemble into hollow cages. We have studied the structural determinants of naturally occurring self-assembling ferritin cages and identified a single mutation that greatly increased the stability of the ferritin cage, as well as dramatically altered the overall structure of the assembly. We have also used the formulation of probabilistic protein design to arrive at novel sequences for α-helical peptides that can adjust their surfaces in accordance to different local environments. This formulation was used to identify a sequence for a peptide designed to self-assemble into a spherical particle with broken symmetry. Taken together, these efforts will lead to an increased understanding of the role of kinetics and structural plasticity in protein-ligand and protein-protein interactions.