Drug combinations are an increasingly favored strategy for increasing therapeutic windows for potential drugs, but enthusiasm for this approach is tempered by concerns that therapeutic synergy will too often be mirrored by synergistic toxicity. Here we review our recent experimental results and numerical simulations that establish the context-specificity of synergistic combinations. Thus systematic testing of chemical combinations in cell-based disease models can preferentially discover synergies with beneficial therapeutic selectivity. For an anti-inflammatory combination, we demonstrate how such selective synergy is achieved through differential expression of its targets in cells associated with therapeutic and toxic effects, and validate the combination's therapeutic relevance in animals. The narrower context specificity of synergistic combinations creates many new opportunities for such therapeutically relevant selectivity, and reinforces the realization that the most useful paradigm for a drug target is often a set of biomolecules that cooperate to produce a therapeutic response with reduced side effects.