Local energetic frustration: Protein evolution, conformational dynamics and design in the age of AI

Proteins operate under competing demands imposed by stability, dynamics, and function, all of which are shaped by evolution. Local energetic frustration provides a quantitative framework to describe how these competing requirements are distributed within the native states of proteins, identifying regions where interactions are optimized and others where energetic conflicts are retained to enable functional behavior. In recent years, the study of local frustration has expanded significantly, driven by the integration of large-scale structural datasets and advances in artificial intelligence methods. Comparative analyses have shown that frustration patterns encode evolutionary pressures across protein families, with minimally frustrated interactions stabilizing structural cores and highly frustrated regions often associated with catalysis, binding, conformational transitions as well as pathogenic phenotypes. At the same time, modern protein language models and structure prediction methods seem to implicitly capture the statistical and structural features underlying frustration, enabling its prediction directly from sequence or structure at proteome scale. These developments suggest that local energetic frustration may be interpreted as an emergent property of the evolutionary information learned by AI models. Here, we review recent advances in the analysis and prediction of local frustration and discuss how this framework could provide mechanistic insights into protein evolution, conformational dynamics, and design. We further argue that incorporating frustration into computational and experimental strategies will be essential to move beyond purely stability-driven approaches toward the rational engineering of functional proteins.