Protein:protein interactions play critical roles in numerous biological processes. Despite the importance of these interactions, single amino acid mutations can enhance or disrupt protein:protein binding affinity, potentially altering biological function and in some cases giving rise to disease. It would be highly useful if we could predict and understand a mutation’s effect on affinity as this could reveal its functional implications, which may help elucidate the underpinnings of disease and accelerate biologics drug discovery. However, existing experimental and computational methods that measure or predict changes in binding affinity have tradeoffs in accuracy, generalizability, and cost of resources. Therefore, there is a need for approaches which are resource-efficient without sacrificing accuracy or generalizability. In this thesis dissertation, I work towards developing such approaches by facilitating improvements in physics-based computational simulations such that they more easily complement experimental techniques. First, I demonstrate the value of integrating molecular dynamics simulations with experimental methods to mechanistically understand how mutations in a disease-relevant protein:protein complex influence binding affinity. Next, I enable advancements in alchemical free energy calculations for predicting mutational effects on binding free energy by contributing to an open-source software package which enables community-wide methods development. Finally, building upon the previous chapter, I provide a deep-dive investigation into the conformational sampling challenges that arise when applying the aforementioned alchemical calculation package to a model protein:protein complex. Taken together, this work illustrates the value and limitations of physics-based simulations for assessing mutational impact on protein:protein interactions, paving the path towards novel applications and further improvements in methodology.