The simplicity of protein sequence-function relationships.

It is widely thought that a protein's function depends on complex interactions among amino acids. If so, it would be virtually impossible to predict the function of new variants, and understanding how proteins work genetically and biochemically would require huge combinatorial experiments. We show that prior studies overestimated complexity because they analyzed sequence-function relationships from the perspective of a single reference genotype and/or misinterpreted global phenotypic nonlinearities as complex amino acid interactions. By developing a new reference-free approach and using it to reanalyze 20 experimental datasets, we show that additive effects and pairwise interactions alone, along with a simple global nonlinearity, explain the vast majority of functional variation. Higher-order interactions are weak or rare, and a minuscule fraction of possible interactions shape each protein's function. Our work reveals that protein sequence-function relationships are surprisingly simple and suggests new strategies that are far more tractable than the massive experiments currently used.