Effective gene therapy must deliver therapeutic genes to disease sites while avoiding healthy tissue. However, engineering targeted gene delivery vectors to ensure exclusive delivery to diseased sites remains a challenge. Adeno-associated virus (AAV) is receiving increasing attention for its potential as a gene delivery vehicle because it offers several advantages: it is considered the safest viral vector, it infects human cells efficiently, and it can be genetically altered to improve therapeutic efficacy. However, even slight modifications to the virus capsid (the outer protein shell covering its genome) lead to unpredictable outcomes. Thus, a governing set of design rules for virus capsid assembly and function is needed to improve future engineering efforts. To this end, this thesis uncovers some of these rules by applying a computational model, often used in protein engineering, to the AAV capsid. A new strategy to improve AAV targeting was also explored by engineering AAVs to sense and become activated by extracellular proteases found in diseased tissues. The specificity of these protease-activatable viruses can be tuned to recognize a variety of protease profiles to treat a multitude of diseases. Design rules for these platform technologies are unveiled through their development and in-depth characterization. We also explore new motifs in the AAV capsid to further our understanding of AAV basic biology. Ultimately, these studies advance our ability to program virus nanoparticles for many biomedical applications.