Newsletter - Edition 26
The Role of Protein Structure in Drug Discovery
Oxford Innovation Society Lecture September 1998
Dr Malcolm Weir, Glaxo Wellcome
The past few years have seen an explosion in the availability of structural information pertaining to drug targets, and the growth of computational chemistry and bioinformatics methods to exploit them; at the same time combinatorial chemistry and screening technologies have greatly advanced, calling into question how much structure-based design input is required to discover small-molecule leads and then optimise their potency and pharmacokinetic properties to produce useful drugs. The central question underlying this debate is "can random synthesis and screening supply us with all the leads we need for easy optimisation?"; simple sums based on screening and synthesis costs set against the vastness of chemical diversity (1018 to 1050 small-molecules, depending on how you calculate it), combined with practical lead discovery experience, suggest the answer currently to be "no", forcing us to make scientific choices of molecule sets for synthesis and bioassay. (Such choices are based on current knowledge, and are thus inclusive of prior serendipitous discoveries, which can be placed within a sound framework and properly exploited.) The figure below illustrates the bases for such design/selection, the most powerful of which is 3D structural data on protein targets or the active conformations of their ligands.
Useful 3D experimental information on target proteins and their ligands comes in several forms, as illustrated in the talk by the examples of X-ray crystallographic structures of thrombin complexes and NMR information on the bound conformation of the E-selectin ligand, sialyl Lewis-X. The synergy between random screening (in this case of natural produce plant extracts) and structural/mechanistic studies is nicely illustrated by the 5,5-translactone series of serine protease inhibitors, which although initially discovered as thrombin inhibitors have since been used as a more general template for drug discovey.
Computational chemistry is employed to exploit structural information and suggest individual compounds or library sets for selection or synthesis. The joining together of de novo design methods with selection of monomers and templates which are amenable to combinatorial chemistry ("structure-based library design") is a promising way of generating novel active molecules, since it counteracts uncertainty in the calculation of binding energies by generation of a focussed set of hundreds or thousands of molecules, and allows the chemist to draw upon complex structural and theoretical information in a quite intuitive way.
When a 3D structure is not available for a target protein, history shows that the familial relationships of the protein can still narrow the chemical space in which we should search for active small-molecules. The first step is to classify the protein by sequence or structural domain, and by its natural ligands; putting the target in context enables retrieval or synthesis of compounds which have an enhanced probability of binding. The common denominator is a structurally conserved (or converged) region of the target family which recognises the complementary small-molecule class. The kinase family illustrates this concept. Clearly such information is relatively soft, and the chemical focus correspondingly widened, but it does allow for the practical application of bioinformatics and homology modelling to lead discovery by linkage to prior small-molecule information, whether from random screening or design.
Finally, although 3D structures of soluble proteins are an ever-increasing and valuable resource, the lack of information on membrane receptors and multidomain proteins continues to frustrate us. Recent advances in cryoelectron microscopy and NMR hold great promise for curing this problem, as do improved expression and purification methods. The ultimate aim must be a complete structural picture of the cell and all its contents! This would truly enable us to analyse protein function, probe ligand binding sites and design drugs.
Newsletter - Edition 26 Contents
- The Role of Protein Structure
- Protein Folding
- New Generation Vaccines
- Maximising the Return from your Research