lecular dynamics simulations have been used to extensively explore the conformational space available to I and II. Six dissimilar conformations of I initiated a total of 130 ps of simulations, and four conformations of II were each subjected to a 20 ps simulation. The lowest energy confor- mations found for each peptide were not in agreement. A new conformation of II was constructed from the exact geometry of the sequence HIS to TRP of the lowest energy conformation of I, which minimized to give a lower energy conformation than had previously been found, and which was in good agreement with I. This conformational space of II has been thoroughly explored by high- and low-tempera- ture dynamics, and by simulated annealing. This has led to a further low energy conformation, which is still in close agreement with that of I. This supports the development of lactam analogues as potent melanotropins.

fits the inverse distance dependence of the potential to an expansion of Gaussian functions, and applies these func- tions to the Carbo similarity index. The approach is com- pared with the numerical grid-based technique of the ASP program using two test systems. In the first investigation a hypoglycemic active ring fragment is compared with a num- ber of related ring analogues. The second investigation in- volves the similarity index optimization of an Np-apomor- phine molecule against a second Np-apomorphine molecule occupying a different region of space. Test results illustrate that for single point calculations little difference exists be- tween the two methods. For similarity optimizations, how- ever, the new Gaussian-based technique exhibits both supe- rior speed (up to two orders of magnitude) and convergence properties. The new approach is thus shown to significantly enhance the potential flexibility of molecular similarity cal- culations.

DESIGNING A SMALL PEPTIDE TOXIN USING MOLECULAR DYNAMICS

Indira Ghosh Astra Research Centre India, P.B. No. 359, 18th Cross Malleswaram, Bangalore 560 003, India

Heat-stable enterotoxin is an 18-mer peptide, and belongs to a family of enterotoxin produced by E. coli. The active fragment is found to be a 13-mer peptide with a sequence of CYS-CYS-GLU-LEU-CYS-CYS-ASN-PRO-ALA-CYS- THR-ALA-CYS. The biologically active compound con- tains three disulfide bridges (between 1 and 6, 2 and 10, and 5 and 13), which give rise to a very compact structure of the molecule to bind at the receptor site. This 13-mer peptide has been modeled using molecular mechanics and dynam- ics. The formation of relaxed disulfide bridges has been used as criteria for determining the feasible conformations of the peptide. The constraining force due to covalent bond formation of disulfides has been found to play an important role in folding the peptide into its active form. This pre- dicted structure has been compared with the crystal structure of a similar toxin produced by a pathogenic strain of E. co/i.’ This method demonstrates the importance of using molecular dynamics simulation for designing disulfide con- taining small molecules.

REFERENCE

1 Shimonishi, Y. et al. J. Biol. Chem. 1991, 266, 5934

THE UTILIZATION OF GAUSSIAN FUNCTIONS FOR THE RAPID EVALUATION OF MOLECULAR SIMILARITY

A.C. Good Physical Chemistry Laboratory, South Parks Road, Oxford OX1 342, UK

An analytic method that permits the comparison of molecu- lar electrostatic potential is introduced. The new technique

MOLECULAR MODELING OF THE HUMAN ERYTHROCYTE GLUCOSE TRANSPORTER

Paul A. Hodgson,* David J. Osguthorpe,* and Geoffrey D. Holmant *Molecular Graphics Unit, School of Chemistry and iDepartment of Biochemistry, University of Bath, Bath, BA2 7AY, UK

We present a model for the three-dimensional structure of the human erythrocyte glucose transporter (GLUTl). The GLUT1 type transporter is thought to be a member of the 12 membrane-spanning segment sugar transport protein super- family. Despite the growing number of sugar transport pro- tein sequences showing the characteristic 12 hydrophobic segments, none has been crystallized to date. Because of the paucity of solved X-ray structures for transmembrane pro- teins there is no real crystallographic template upon which to base a model sugar transport protein. Therefore, our model is based upon evidence from many sources. Inferences rele- vant to the three-dimensional structure of GLUT1 drawn from such data have enabled us to construct a model consist- ing of a bundle of 12 membrane-spanning a-helices. Each of the two transmembrane domains of the protein-helices l-6 (NH, domain) and 7-12 (COOH domain)-is a left- handed six-helical bundle, the helices being joined by rela- tively short linkers (8-10 residues). Putative interdomain repeats have allowed us to model the NH, domain on the COOH domain, for which more data exists. The two transmembrane domains are joined by a cytoplasmic do- main. This domain, predicted to consist of an antiparallel (Y- helical hairpin pair, lies in the plane of the membrane between the sixth and seventh membrane-spanning helices.

AN ACCURATE SIMULATION OF THE PROTEIN STREPTOMYCES GRISEUS PROTEASE A

David H. Kitson,* Franc Avbelj,t# John Moult,# Dzung T. Nguyen,** John Mertz,$ D. Hadzi,? and Arnold T. Hag- ler**

62 J. Mol. Graphics, 1993, Vol. 11, March