Applications of artificial intelligence in drug design

Abstract

Computer-aided drug design is a rapidly growing area of research. The design process can proceed from two angles: either the three-dimensional structure of the biological target is known, or it is unknown. Thus the area of computer-aided drug design can be separated into a number of problems. One problem is determining the structure of a biomolecule from experimental data, as is done in chapter 2 for colominic acid polylactone. These studies determined that there are two helical structures consistent with spectral data A second problem is designing a ligand complementary to the three-dimensional structure of the target. Chapters 3 and 4 describe studies leading to the design and evaluation of neuraminidase inhibitors. These studies indicate that several inhibitors studied are competitive inhibitors of the enzyme with better binding affinities than the natural ligand. The final,and potentially most difficult problem, is to infer features about the biological target from compounds known to bind to that target. Chapters 5 and 6 describe model studies and implementation of CLEW, a program to learn rules relating structural features to biological function. Results indicate that learning based on topological features is a useful first iteration in determining the pharmacophore, or three-dimensional arrangement of functionality required for biological activity.