Originally, rotamers wre defined as side-chain torsion (X-angle) combinations corresponding to the local minima of potential energy (van-der-Waals and torsion terms) for the side-chain of a terminally blocked amino acid. If at least one X-angle differed by more than 20° from that of a rotamer, the side-chain was considered as deviant both from energetic (increase in potential energy of no less than 1 to 2 kcal/mol) and geometric (precision of atom positioning is worse than 0.5 Å) aspects. In this work the distribution of side-chain conformations in protein crystal structures is analysed. Large deviations from rotamerie X-values occur systematically and cannot be attributed merely to errors in crystal structure determination. The “rotamericity” (the fraction of residues within ±20° of the, X-angles of a rotamer) not only remains substantially below 100% (70 to 95% for various amino acids) with improving crystallographic resolution but actually decreases for 8 out of 17 amino acid types after a critical resolution limit is crossed. This effect has been observed for external as well as for internal residues. The set of amino acid side-chain conformations in globular proteins cannot be considered as normally distributed around some rotamer points. Outliers occur systematically. The rotamericity of an amino acid depends essentially on the different environments the amino acid meets in real protein structures. Factors such as the backbone torsion angles of the residue itself, the secondary structure and tertiary contacts influence the rotamericity. The deviations in regions of regular main-chain, structure from the average g-:t:g+ relationship in the X1 angle become much more evident if, in addition to the typical secondary structure assignments, the actual backbone torsion angles of the residue are taken into account. In α-helices the t:g+ distribution in the X1-angle correlates with physical properties describing volume, extension and flexibility of the side-chain. In α-strands the factors influencing the t:g+ distribution in the X1-angle are the polarity and hydrophobicity of the side-chain. Nevertheless, a considerable number of residues do not comply with the statistical preferences observed for tite side-chain conformation. Large deviations from the rotamer values are observed especially in cases when normally advantageous X1-values are not allowed and adjustments in X2 become nenessary to accommodate the side-chain. For example, in the case of phenylalanine in very regular α-helices, the trans-conformation is strongly preferred in the X1-angle with X2 near 90°; yet in about one-third of all phenylalanine residues this conformation is not allowed because of hard-sphere atomic clashes, almost exclusively due to tertiary contacts with other side-chains. The corresponding phenylalanine residues adopt a X1-value in the g+-region while the X2-values are nearly uniformly distributed from 0° to 180°. Hence, in these cases the main-chain conformation is not of predictive value for the side-chain conformation, and the X-angle selection is based upon specific and local tertiary interactions. A substantial number of amino acid side-chains are under strain. Apparently, the relaxation of some side-chains is sacrificed for a global minimum of free energy to be achieved at a higher organizational level involving the protein fold. A similar observation is also known from the main-chain conformation. The results are relevant for the development of algorithms that model proteins using restricted scanning of the torsion angle space. © 1993 Academic Press, Inc.
