Researchers have developed a method to predict the 3D structure of proteins based on evolutionary sequence variation. By analyzing the co-evolution of amino acid residues in homologous proteins, they inferred residue pair couplings that accurately predict residue-residue proximity in folded structures. Using these couplings, they computed all-atom 3D structures of 15 test proteins, including a G-protein coupled receptor, with high accuracy. The method, called EVfold, achieved a Cα-RMSD error of 2.7–4.8 Å for at least two-thirds of the protein, without using homology modeling or known structures. This approach leverages statistical models to infer residue-residue contacts from multiple sequence alignments, revealing essential interactions that constrain protein evolution. The results suggest that evolutionary sequence variation contains sufficient information to predict protein structures accurately, which could advance protein and drug design, and the identification of functional genetic variants. The study highlights the potential of using evolutionary data to infer protein structures, offering new insights into protein folding and evolution.Researchers have developed a method to predict the 3D structure of proteins based on evolutionary sequence variation. By analyzing the co-evolution of amino acid residues in homologous proteins, they inferred residue pair couplings that accurately predict residue-residue proximity in folded structures. Using these couplings, they computed all-atom 3D structures of 15 test proteins, including a G-protein coupled receptor, with high accuracy. The method, called EVfold, achieved a Cα-RMSD error of 2.7–4.8 Å for at least two-thirds of the protein, without using homology modeling or known structures. This approach leverages statistical models to infer residue-residue contacts from multiple sequence alignments, revealing essential interactions that constrain protein evolution. The results suggest that evolutionary sequence variation contains sufficient information to predict protein structures accurately, which could advance protein and drug design, and the identification of functional genetic variants. The study highlights the potential of using evolutionary data to infer protein structures, offering new insights into protein folding and evolution.