It remains extraordinarily difficult to identify causal genes in most genetic diseases, in particular highly polygenic disorders, such as, for example, coronary artery disease, diabetes, and autism, for which current approaches are most limited. I'll discuss our progress on this problem, which builds upon the tendency for many genetic traits and disorders to arise from the dysfunction of specific biochemical machines, such as complexes or systems of proteins. We use a combination of computational and experimental strategies to discover disease-relevant systems of proteins. One approach for identifying candidate genes for diseases, based on identifying surprising connections between human diseases and apparently unrelated traits in distant organisms, led us to discover that yeast cells can serve as a model for certain defects in blood vessel development. This in turn allowed us to repurpose a common antifungal drug as a potential chemotherapy. Similar analyses revealed a plant model for human deafness, and led us to launch major efforts in systematically humanizing yeast cells and in applying high-throughput protein mass spectrometry to measure the physical associations among thousands of proteins and their evolutionary conservation across major branches of life. Studies such as these reveal the evolutionary basis for traits and diseases and give us insights into the basic physical infrastructure of cells.