The authors performed a systematic, large-scale analysis to create a phenome-interactome network of human protein complexes involved in various diseases. They integrated quality-controlled protein interactions with a validated phenotype similarity score to identify previously unknown complexes associated with diseases. Using a Bayesian predictor, they correctly ranked known disease-causing proteins in 298 out of 669 linkage intervals and identified novel candidates for 870 intervals without identified disease-causing genes. The network includes 506 disease-associated protein complexes, which reveal functional relationships between disease-promoting genes. The study demonstrates the value of integrating protein interaction data from model organisms and human phenotype data to prioritize positional candidates for disease genes. The authors also present four case studies to illustrate how the complexes can generate novel hypotheses for validation.The authors performed a systematic, large-scale analysis to create a phenome-interactome network of human protein complexes involved in various diseases. They integrated quality-controlled protein interactions with a validated phenotype similarity score to identify previously unknown complexes associated with diseases. Using a Bayesian predictor, they correctly ranked known disease-causing proteins in 298 out of 669 linkage intervals and identified novel candidates for 870 intervals without identified disease-causing genes. The network includes 506 disease-associated protein complexes, which reveal functional relationships between disease-promoting genes. The study demonstrates the value of integrating protein interaction data from model organisms and human phenotype data to prioritize positional candidates for disease genes. The authors also present four case studies to illustrate how the complexes can generate novel hypotheses for validation.