An astounding 90% of hits identified in drug screening will fail during lead optimization, according to Herman Verheij, head of the computational chemistry team at Pyxis Discovery. “Several top ten pharma companies have thrown away, on aggregate, more than two million compounds because of quality issues and bad physicochemical properties” he adds. What can be done to improve the quality and potential of the drug candidates emerging from HTS campaigns? One trend sees pharma companies moving away from screening huge, diverse compound libraries to searching for leads in smaller libraries and analogue collections. Such libraries are the focus of more intense scrutiny, biological design, and physicochemical tinkering upfront, even before they are used in a primary screen.

The synthesis and screening of fragment libraries is at the leading edge of this trend toward more design-conscious compound library synthesis. Fragment-based approaches emerged from the understanding that a molecular weight >500 Daltons typically correlates with poorer oral bioavailability. Furthermore, drug compounds tend to be even larger than the leads from which they were derived as a result of lead optimization efforts. In addition, a renewed interest in mimicking molecules found in nature and in leveraging the success of marketed drug compounds, which both tend to be large, complex, and highly lipophilic chemical entities, led to efforts to identify the individual components of these molecules and isolate those responsible for the desired biological activity that would then be amenable to medicinal chemistry.

Using natural products as the basis for medicinal chemistry presents several challenges. “Their diversity is enormous; they often have multiple chiral centers and it is difficult to know which are important for activity,” says John Barker, a structural biologist at Evotec.

A recent poster from Evotec describes a rapid fragment-to-lead process the combines fragment library synthesis; x-ray crystallography to analyze target-fragment binding,including visualization of co-crystal structures in which more than one fragment binds simultaneously to two different target sites; medicinal and parallel chemistry to link and modify fragments to enhance potency and selectivity; and fluorescence-based high-throughput fragment screening (HTFS) using the company’s EVOscreen Mark II.

Medicinal chemists at Evotec select fragments for the library based on their suitability for analog synthesis and characteristics associated with drug-like scaffolds, including absence of toxicophores; logP (lipophilicity) <3; solubility > 10^-3M; rotatable bonds <5; hydrogen-bond acceptors (HBA) <4 and H-bond donors (HGC) < 3; and topological polar surface area < 70 Ų. Following crystal structure analysis to define key interactions and the potential for fragment evolution or linking, chemists then define synthetic strategies for generating analogues to the fragments of interest. Synthesis strategies are devised for generating a compound with optimal biophysical and binding properties.

Barker, a structural biologist at Evotec, describes the fragments as “sticky molecules but not promiscuous binders.” The company’s fragment library will soon grow from 5,000 to 20,000 unique chemical entities. “Each is hand-picked by a panel of medicinal chemists.” Of particular importance is a fragment’s solubility — to minimize aggregation or precipitation — a critical attribute for both high concentration screening and crystallography.

These same trends figure prominently in the strategy Tranzyme Pharma has adopted to design synthetic libraries of orally bioavailable macrocyclic compounds capable of targeting a broad spectrum of drug target classes, including GPCRs, protein kinases, protein-protein interactions, and ion channels. By mimicking the structures of macrocyclic compounds commonly found in nature that have known drug-like activity, including peptides and macrolide antibiotics, Tranzyme’s MacrocyclicTemplate Chemistry (MATCH) combines multiple amino-acid based recognition elements that bind with high affinity and selectivity to a target’s active site with a unique non-amino acid component to control compound topology. This component limits a molecule’s flexibility and fixes the structure in 3D space to limit its conformational populations.

As amino acid side chains are intimately involved in protein-protein or peptide-protein interactions in nature, starting with these recognition elements and putting them in a macrocyclic framework yields “compounds with intrinsic biological relevance,” says Mark Peterson, vice president of IP and operations at Tranzyme. The company recently received European patent protection for its core macrocyclic chemistry and, according to Peterson, has also received a notice of allowance from the U.S. Patent Office.

The company designed a group of “tethers” — the key proprietary component of its compounds. These tethers link two ends of a molecule together, creating conformational restriction. By fixing the 3D shape of the molecule, the tether element maximizes the compound’s potency and selectivity and also “allows us to tailor the compound’s pharmacodynamic and pharmacokinetic properties,” Peterson says. “Cyclization stabilizes the molecule as well, preventing rapid metabolic breakdown.”

Tranzyme’s combinatorial synthesis strategy involves combining amino acid recognition elements representing multiple pharmacophores with numerous tethers to yield libraries of macrocyclic compounds with molecular weights in the range of 400 to 550 Daltons . The initial synthetic library for screening contains approximately contains 25,000 structures.

“Natural product-like synthetic libraries are only now being more widely appreciated and developed,” asserts Peterson. Tranzyme’s progress in drug discovery, in the area of gastrointestinal and metabolic diseases, and the entry of its lead compound, TZP-101, into Phase II trials, validates the potential of this technology for generating novel drug-like molecules with oral bioavailability.

Later this year, Pyxis Discovery will add fragment libraries to its current range of Smart Libraries. Fragment-based screening is “really hot at the moment,” says Herman Verheij. Conventional biochemical assays will typically identify compounds consisting of two or three already linked fragments that each bind weakly to different sites on a drug target, he explains. The company identifies individual fragments that each bind strongly to these sites and combines them to yield high affinity compounds.

Library design at Pyxis is now also focusing on known drugs and natural compounds, by looking for substructures in compounds from those two classes of molecules that have proven biological merit and that are amenable to medicinal chemistry. Verheij describes the ability to identify the core elements in natural or drug compounds that are responsible for their biological activity and are known to be well accepted by the human body as the “Holy Grail” of library design.

Pyxis’s newest Smart Library, Source 10.06, contains 2,964 chemically diverse molecules enriched with heterocyclic or aromatic rings and linkers known to be present in biologically active molecules. Whereas most Smart Libraries emphasize diversity, others are focused towards specific target classes, including GPCRs, kinases, and proteases.

Copyright 2007, Cambridge Healthtech Institute. All Rights Reserved.