Back propagation neural networks is a new technology useful for modeling nonlinear functions of several variables. This paper explores their applications in the field of quantitative structure-activity relationships. In particular, their ability to fit biological activity surfaces, predict activity, and determine the "functional forms" of its dependence on physical properties is compared to well-established methods in the field. A dataset of 256 5-phenyl-3,4-diamino-6,6-dimethyldihydrotriazines that inhibit dihydrofolate reductase enzyme is used as a basis for comparison. It is found that neural networks lead to enhanced surface fits and predictions relative to standard regression methods. Moreover, they circumvent the need for ad hoc indicator variables, which account for a significant part of the variance in linear regression models. Additionally, they lead to the elucidation of nonlinear and "cross-products" effects that correspond to trade-offs between physical properties in their effect on biological activity. This is the first demonstration of the latter two findings. On the other hand, due to the complexity of the resulting models, an understanding of the local, but not the global, structure-activity relationships is possible. The latter must await further developments. Furthermore, the longer computational time required to train the networks is somewhat inconveniencing, although not restrictive.

We have performed molecular dynamics calculations for liquid water using the revised central force model potential truncated at various distances and using two sizes of system, in order to study the effect of system size and range of the potential on the calculated thermodynamic and structural properties and to compare the results with those obtained by Ewald summation. All calculations were performed for a cubic system using periodic boundary conditions. Provided the side of the cube is equal to or greater than twice the range of the potential, the thermodynamic properties and distribution functions, including an orientational distribution function, are insensitive to the size of the system, for fixed range. Provided the range of the potential is equal to or greater than 6 Å, the thermodynamic energy and pressure are only slightly dependent on the range of the potential, to an extent that is, however, larger than that observed for dipolar hard spheres and the Stockmayer potential which do not have the tetrahedral structure similar to water. For potentials with ranges of 6 Å or greater, the atom–atom distribution functions are very insensitive to the range, but the orientational correlations are very sensitive to the range, as had been observed many times in studies of dielectric properties of simulated fluids. A potential with a range of 6 Å has thermodynamic and structural properties very similar to those of a longer ranged potential and similar to those obtained by Ewald summation. Use of such a model, which correctly describes interactions between nearest neighbors and next nearest neighbors but has no longer ranged forces, lead to significant increases in the speed of simulations.

By using a homology-based bioinformatics approach, a structural model of the vaccinia virus (VV) I7L proteinase was developed. A unique chemical library of ϳ51,000 compounds was computationally queried to identify potential active site inhibitors. The resulting biased subset of compounds was assayed for both toxicity and the ability to inhibit the growth of VV in tissue culture cells. A family of chemotypically related compounds was found which exhibits selective activity against orthopoxviruses, inhibiting VV with 50% inhibitory concentrations of 3 to 12 M. These compounds exhibited no significant cytotoxicity in the four cell lines tested and did not inhibit the growth of other organisms such as Saccharomyces cerevisiae, Pseudomonas aeruginosa, adenovirus, or encephalomyocarditis virus. Phenotypic analyses of virus-infected cells were conducted in the presence of active compounds to verify that the correct biochemical step (I7L-mediated core protein processing) was being inhibited. Electron microscopy of compound-treated VV-infected cells indicated a block in morphogenesis. Compound-resistant viruses were generated and resistance was mapped to the I7L open reading frame. Transient expression with the mutant I7L gene rescued the ability of wild-type virus to replicate in the presence of compound, indicating that this is the only gene necessary for resistance. This novel class of inhibitors has potential for development as an efficient antiviral drug against pathogenic orthopoxviruses, including smallpox.

No abstract

Using a data set comprised of literature compounds and structure-activity data for cyclin dependent kinase 2, several pharmacophore hypotheses were generated using Catalyst and evaluated using several criteria. The two best were used in retrospective searches of 10 three-dimensional databases containing over 1,000,000 proprietary compounds. The results were then analyzed for the efficiency with which the hypotheses performed in the areas of compound prioritization, library prioritization, and library design. First as a test of their compound prioritization capabilities, the pharmacophore models were used to search combinatorial libraries that were known to contain CDK active compounds to see if the pharmacophore models could selectively choose the active compounds over the inactive compounds. Second as a test of their utility in library design again the pharmacophore models were used to search the active combinatorial libraries to see if the key synthons were over represented in the hits from the pharmacophore searches. Finally as a test of their ability to prioritize combinatorial libraries, several inactive libraries were searched in addition to the active libraries in order to see if the active libraries produced significantly more hits than the inactive libraries. For this study the pharmacophore models showed potential in all three areas. For compound prioritization, one of the models selected active compounds at a rate nearly 11 times that of random compound selection though in other cases models missed the active compounds entirely. For library design, most of the key fragments were over represented in the hits from at least one of the searches though again some key fragments were missed. Finally, for library prioritization, the two active libraries both produced a significant number of hits with both pharmacophore models, whereas none of the eight inactive libraries produced a significant number of hits for both models.