All pharmaceutical researchers know the feeling. Somewhere out there must be that elusive molecule — one that will inhibit this enzyme or activate that receptor in the way they want, and without causing unwanted side-effects. But finding it is another matter. For small-molecule drugs — the mainstay of the pharmaceutical industry — time-consuming and expensive screening is needed to pick out promising candidates from the vast number of natural and synthetic compounds available. Testing large numbers of compounds to see if they produce an appropriate biochemical or cellular effect is usually one of the first steps in the drug-discovery pathway, and ways of making this screening faster, more effective and less expensive are in continual development.

A positive response in a first round of screening in a biochemical assay identifies the primary 'hit' compounds. These molecules then go into more screens to see if they have physicochemical and pharmacological properties that are not too incompatible with making a drug — if it passes this filter, a hit becomes a 'lead'. Lead compounds then undergo further rounds of chemical refinement and biological screening before finally entering clinical testing. With a good deal of luck, your lead might eventually be approved as a drug 12–15 years after testing began.

R&D spend versus new drug launches by the top 20 pharma companies.

But all is not quite as it should be in the drug industry. Estimates vary, but in general analysts agree that each major pharma company needs to launch three or four new products a year in order to sustain the present level of growth. A glance at the chart on the right shows that productivity over the past few years has been well below this level, with the top 20 pharma companies averaging just over one new launch a year.

Increasing the number of leads is thus high on the drug-discovery agenda. Foreseeing the coming deficit, companies implemented a number of strategies in the late 1980s intended to do this. Combinatorial chemistry was used to generate larger libraries of compounds for testing, and high-throughput technology, including increasing miniaturization and automation, was deployed to screen these libraries more rapidly. But despite tremendous advances in all aspects of the screening process, chiefly the increased use of automation, these improvements did not bring about the expected rise in productivity, and the industry's drug pipelines still look decidely thin.

Advocates of high-throughput screening claim that the technique is still in its infancy. “High-throughput screening is not 100,000 tests a day, it's 100,000 tests every day,” says Richard Archer, chief executive of The Automation Partnership in Royston, UK, a company that makes automated equipment for screening. “You can't say that automated high-throughput screening doesn't work, because nobody is doing it yet.”

But all now agree that for screening, quality is more important than quantity. Throughout the industry the emphasis is shifting — from screening the greatest number of compounds as quickly as possible, to making sure that high-quality compounds are going into robust and reproducible assays, and to understanding potential targets better (see "Getting to know the family").

Getting organized

The basic workhorse of screening is the microtitre plate, and a number of companies have developed robust, automated screening platforms based on this format. Initially, the plates featured 96 wells, which allowed the same number of different molecules to be screened for a given activity. More recently, the miniaturization of the simpler types of assay has seen the 384-well plate become standard, leaving those with 96 wells to be used in more complicated assays.

Portable screening: Amersham's LEADseeker Credit: AMERSHAM BIOSCIENCES

Plates with an even higher capacity (1,536 and even 3,456 wells) are used more rarely because the problems associated with handling minute volumes can add significantly to the cost of the assay (see "How small should you go?", page 457). Amersham Biosciences, an equipment manufacturer based in Piscataway, New Jersey, is one of several companies producing the new generation of robust screening platforms. Its LEADseeker, for example, is designed for decentralized primary screening. It uses imaging technology based on a charge-coupled device that detects fluorescence and luminescence, and allows a whole 96-, 384- or 1,536-well plate to be read at a time.

Amersham sees the LEADseeker as a step on from earlier technologies based on tracking radiolabelled samples, such as scintillation proximity assays. Indeed, a general feature of the new generation of equipment is that it uses fluorescence-based assays. These have high signal-to-noise ratios, and therefore offer higher-quality data compared with radioactivity-based assays — so much so that in many cases signal detection is so clear there is no need to do replicate wells.

But supplying the equipment is just one part of the equation. Manufacturers also recognize that organizing the high-throughput laboratory's workflow is equally important (see "Automating the screening process", page 453). “Different companies want different things,” says Mike Evans, vice-president for bioassays at Amersham. “Some want 'turnkey solutions', whereas others want to mix and match with piecemeal technology.”

Making products tailored to an individual user's requirements is also becoming a common theme among providers of software and bioinformatics solutions for screening. Here, the main cry from the industry is for data-handling packages that conform to common standards so that they can be interfaced with existing systems. “In the past, software companies were sometimes guilty of trying to impose their own standards on the industry,” says Scott Kahn, a senior vice-president at Accelrys, a software manufacturer in San Diego, California.

Accelrys is one of several companies that favour the development of generally recognized external, non-proprietary standards. The company uses Microsoft standards, as does its main competitor, Spotfire in Somerville, Massachusetts. This offers the end-user additional benefits. “The fact that both companies are developing to a common standard means that although they're competing head-to-head, users can integrate their products however they wish,” says Kahn.

The shape of things to come

Libraries of small-molecule compounds are the raw material that goes into the primary screens. Although there is general agreement about how assay platforms should be developing, there seems to be little consensus about the shape of the ideal compound library. Opinions vary on how big a library should be, and how companies should design, store and handle its contents.

One idea that is exciting interest is to profile and filter compounds for drug-like properties such as solubility and lipophilicity before they ever get into the library. This should give medicinal chemists an easier time by ensuring that lead compounds need less refinement to turn them into drugs.

Companies such as Argenta Discovery, a medicinal chemistry design and screening company based in Harlow, UK, are now screening compounds for a range of drug-like behaviours before they enter the company's libraries. Chris Newton, chief scientific officer at Argenta, describes the profiling as “multi-parametric optimization”.

Namchuk: screens are experiments.

Early whittling away of compounds with undesirable properties can also be done by computer, and in silico screening for 'drug-likeness' is a central component of the 'virtual-screening' strategies of companies such as Argenta, De Novo Pharmaceuticals in Cambridge, UK, and Vertex Pharmaceuticals in Cambridge, Massachusetts. “We're trying to encode the common sense of medicinal chemists into the computer,” says Mark Namchuk, head of high-throughput screening at Vertex. The company uses a proprietary program called REOS (rapid elimination of swill) to eliminate non-drug-like molecules before compounds make it through to the primary screen.

It is too early to judge the success of virtual-screening programmes, but two independent teams of researchers, from Merck laboratories in Rahway, New Jersey, and from Brian Shoichet's group at Northwestern University, have shown structure-based computational docking used as a filter can hugely enrich the hit rate compared with random screening.

Compounds on display

One area of chemical screening where the drive towards automation has been somewhat weak is compound handling.

The preparation of microtitre plates — placing the various compounds into their appropriate wells ready for screening — is still relatively slow.

Graffinity Pharmaceuticals, a drug-discovery company based in Heidelberg, Germany, has come up with an alternative strategy. It sprays 10,000 compounds as spots onto a 'chip', and their affinity for a target protein can be read simultaneously by an imager based on the surface plasmon resonance method developed by equipment manufacturer Biacore in Uppsala, Sweden (see "Fragmenting the problem", page 459).

Graffinity's early microarrays were made up of binary combinations of monomers using amide coupling, as these are easy to make and can rapidly generate a large library of compounds. The company now has a more diverse library of 70,000 compounds presented on microarrays. These can be screened against a protein target in a day, requiring just 5 mg of protein.

This microarray platform generates a relatively high number of hits, but many of them will be for compounds with similar structures, because the screen picks up the activities of the monomer building blocks as well as the binary combinations.

High-content screening

The amount of information that can be gleaned from a screen can be increased by using cell-based systems. Screens such as those offered by Amersham Biosciences, Evotec OAI in Hamburg, Germany, and Vertex Pharmaceuticals in San Diego, California, allow complex biological data on lead-compound behaviour to be collected.

“Although the industry has been doing in vitro assays for a long time, there is a big increase in complexity when you start thinking about using whole cells,” says John Anson, vice-president of systems development at Amersham. For instance, instead of just measuring the binding of a ligand to a receptor in vitro, you might now need to track the movement of a labelled molecule from the cytoplasm to the nucleus. Researchers are also beginning to measure more than one event at a time, for instance by using two different reporter molecules, and this is adding to the complexity.

The increased intricacy of assay systems is changing perceptions of the screening process. “The ability to track the internalization or translocation of a cellular component allows you to think more deeply about what you want to get out of a screen,” says Paul Negulescu, vice-president of discovery biology at Vertex.

Although most researchers would admit that a degree of serendipity operates in screening for hits and leads, most screens are hypothesis-driven, using assays designed to test the effects of compounds on a particular protein target. But CombinatoRx, a two-year-old company based in Cambridge, Massachusetts, has taken a very different approach. It screens binary combinations of existing drugs to see whether drugs that have known effects when acting singly might have different, unexpected, effects when used in combination.

Double value

Alex Borisy: CombinatoRx's binary screens pick up complexity. Credit: COMBINATORX

Most drugs do not, in fact, target single proteins, explains the company's chief executive, Alexis Borisy. Instead they interact with a number of targets at a variety of potencies. “Recognizing the inherent complexity of biological systems, we want drugs that will interact with multiple points in a pathway, rather than the 'sledgehammer' strategy of affecting just one key protein,” he says, reversing the usual mantra that drugs should be as selective as possible.

The data generated by CombinatoRx's screens are built into 'interaction spaces' to illustrate the dose–response relationship of the two drugs in combination. At present, the company has a screening library of 12.5 million binary combinations. And because all these molecules have already been approved by the US Food and Drug Administration, and are mostly off-patent, it should be possible rapidly to develop any hits for further testing. CombinatoRx plans to start clinical trials on its first sets of binary combinations later this year.

For all these new approaches to screening, the number of new compounds entering clinical trials in the coming years will be the ultimate measure of their success.