Some of the parameters that are used in the computer program ECEPP (Empirical Conformational Energy Program for Peptides) to describe the geometry of amino acid residues and the potential energy of interactions have been updated. The changes are based on recently available experimental information. The most signifcant changes improve the geometry and the interactions of prolyl and hydroxyprolyl residues, on the basis of crystallographic structural data. The structure of the pyrrolidine ring has been revised to correspond to the experimentally determined extent of out-of-plane puckering of the five-membered ring. The geometry of the peptide group preceding a Pro residue has also been altered. The parameters for nonbonded interactions involving the C6 and H* atoms of Pro and Hyp have been modified. Use of the revised parameters provides improvements in the computed minimum-energy conformations of peptides containing the Pro-Pro and Ala-Pro sequences. In particular, it is demonstrated that an a-helix-like conformation of a residue preceding Pro is now only of moderately high energy, and thus it is an accessible state. This result corroborates the observed occurrence of Pro residues in kinked a-helices in globular proteins. The structure of the poly(G1y-PrePro) triple helix, a computational model for collagen structure, has been recomputed. The validity of previous computations for this model structure has been confirmed. The refinement of the computed interactions has provided a new general model structure to be used for future computations on collagen-like polypeptides.

An algorithm for the rapid analytical determination of the accessible surface areas of solute molecules is described. The accessible surface areas as well as the derivatives with respect to the Cartesian coordinates of the atoms are computed by a program called "MSEED," which is based in part on Connolly's analytical formulas for determining surface area. Comparisons of the CPU time required for MSEED, Connolly's numerical algorithm DOT, and a program for surface area determination (ANA) based on Connolly's analytical algorithm, are presented. MSEED is shown to be as much as 70 times faster than ANA and up to 11 times faster than DOT for several proteins. The greater speed of MSEED is achieved partially because nonproductive computation of the surface areas of internal atoms is avoided. A sample minimization of an energy function, which included a term for hydration, was carried out on MET-enkephalin using MSEED to compute the solvent-accessible surface area and its derivatives. The potential employed was ECEPPiZ plus an empirical potential for solvation based on the solvent-accessible surface area of the peptide. The CPU time required for 150 steps of minimization with the potential that included solvation was approximately twice as great as the CPU time required for 150 steps of minimization with the ECEPP/2 potential only.

A preliminary three-dimensional structure of angiogenin has been computed, based on its homology to bovine pancreatic ribonuclease A. A standard-geometry structure of ribonuclease was first obtained from its x-ray coordinates. The fit of the backbone of angiogenin to that of ribonuclease was then optimized by taking account of amino acid deletions and by minimizing its conformational energy-plus-a-penalty distance function constraining its backbone to that of ribonuclease. Side-chain and backbone dihedral angles were allowed to vary throughout the cycles of energy minimization. In the last stages of minimization, the penalty distance function was removed. A low-energy structure resembling ribonuclease was obtained.Angiogenin is a basic single-chain protein of 123 amino acid residues that induces in vivo angiogenesis, the formation of blood vessels and a vascular system (1-3). The protein has been isolated in pure form from the serum-free conditioned medium of the human colorectal adenocarcinoma cell line HT-29. Both the amino acid sequence (2) and the corresponding DNA sequence (3) have been determined, showing it to have about a 35% homology with several mammalian pancreatic ribonucleases. Since the structure of bovine pancreatic ribonuclease is known (4, 5), the structure of angiogenin can be computed by minimization of its conformational energy on the assumption that it has a three-dimensional backbone structure similar to that of ribonuclease. This approach is used here to compute a preliminary three-dimensional structure of the whole angiogenin molecule. A similar procedure has been used to compute the structure of bovine a-lactalbumin (6) from that of hen egg-white lysozyme (7) and those of several snake venom inhibitors (8) from that of bovine pancreatic trypsin inhibitor (9). METHODOLOGY AND RESULTSOptimized Structure of Ribonuclease. First, a model of ribonuclease (RNase) having the standard geometry (bond lengths and bond angles) of the ECEPP (empirical conformational energy program for peptides) algorithm (10, 11) was computed from the x-ray coordinates (4, 5). The objective was to obtain an energy-minimized, standard-geometry structure for ribonuclease that would superimpose well on the known 2.0-A resolution x-ray structure and be free of any high-energy atomic overlaps. The coordinates from this standard-geometry model could then be used as input to provide a starting conformation for the backbone of angiogenin, as well as for side chains that are conserved between the two molecules. The degrees of freedom in this type of model are dihedral angles, with bond lengths and bond angles fixed at values obtained from high-resolution crystal structures of small molecules. For those dihedral angles that could be defined by the positions of four heavy atoms, initial values were computed from the neutron-and x-ray-derived Cartesian coordinates (4, 5). For the initial ECEPP model, all peptide bond dihedral angles, cl, were fixed at 1800 with the exception of those preceding proline residues, which wer...