Molecular docking: Theory and application to recognition and inhibitor design

Predicting the favourable co-positioning of two interacting molecules is a necessary first step towards predicting function or designing inhibitors based on the structures of biological receptors. In the following chapters I take up some of the theoretical and practical aspects of molecular docking, often involving the use and elaboration of the computer program DOCK.

In Chapter one, I use DOCK and molecular mechanics to predict a ternary complex of Thymidylate Synthase (TS) with its two natural substrates, dUMP and CH$\sb2$-H$\sb4$folate. My predictions are testable by site-directed mutagenesis, crystallography and enzymology.

I develop new docking algorithms in chapter two. The new algorithms improve the efficiency and accuracy of the docking method. I use molecular organization and sampling techniques to remove the exponential time dependence on molecular size in docking calculations. The new techniques allow me to study systems that were prohibitively large for the original method, including 7 complexes where the ligand is itself a protein. In all cases, the new algorithms successfully reproduces the experimentally determined configurations.

These new algorithms are used in chapter three to address "the protein docking problem." I accurately regenerate the structures of three known protein complexes, using both the bound and unbound conformations of the interacting molecules. I also find geometries that did not resemble the crystal structure. Simple complementarity methods (surface area burial, solvation free energy, packing, electrostatic interaction energy and mechanism-based filtering) can not distinguish between 'native' and 'non-native' complexes. Energy minimization is more reliable, though the energy differences are surprisingly small. The regeneration of the crystallographic configurations using the unbound conformations suggests that DOCK will be useful in predicting the structures of unsolved complexes.

In chapter four, I screen a structural database for chemicals complementary to TS using DOCK. Besides retrieving the natural substrate and known inhibitors, I find molecules previously unknown to bind to TS. I test several of these and find two different classes of novel inhibitors. The crystallographic solution of complexes of several of these inhibitors allow us to test our models to atomic resolution. These structures suggest new ligands with improved binding.