Patrick McGee, Senior Editor
Some say high-throughput screening has not delivered on its promise, that the capsule is half empty, but proponents of HTS see the capsule as half full.
A few years ago, articles began appearing in the mainstream press, most notably the Wall Street Journal and New York Times, advancing the idea that the considerable
![]() |
investments made by the pharmaceutical and biotechnology industries in high-throughput screening and combinatorial chemistry in the late 1980s through the 1990s were folly. The articles argued that the billions invested had not delivered the rejuvenated product pipelines promised by their acolytes. The accusations sparked a great deal of anger, and soul searching, in the industry.
While some agree with the charges, others argue they are based on a fundamental misunderstanding of the drug discovery and development process. The hype that surrounded combinatorial chemistry, and especially high-throughput screening (HTS), does not mesh with the long chain that is modern drug development. They argue that the HTS investments made over the last decade and a half are just now maturing and will have more and more of an impact over the next several years. “Today, high-throughput screening is more integrated into the drug discovery paradigm, but it’s not the panacea, it’s not the answer. Drug discovery is very complex and difficult,” says David Burns, PhD, manager of the HTS group at Abbott Laboratories, Abbott Park, Ill.
Philip Tagari, PhD, director of research at Amgen Inc., Thousand Oaks, Calif., says many do not understand the fact that a large sample collection does not necessarily translate into a high quality and genuinely chemically diverse library, particularly in the sense of bioactivity, of molecules to screen. “The ‘hit’ that in itself, or with minimal modification, becomes a drug, is still a great rarity,” he says, because available chemical collections have many liabilities. “However, chemical sciences are advancing rapidly, which will result in the availability of more genuinely drug-like libraries and, thus, a vastly improved chance of finding those one-in-a-million molecules.”
Jeff Pasley, PhD, vice president for HTS at Wyeth Laboratories, Collegeville, Pa., believes the investment in HTS is paying off. He cites the data of Sandra Fox of High Tech Business Decisions, Moraga, Calif., who has conducted research into the impact of HTS based on interviews with anonymous sources in the industry. A study she released in 2005 found that as of October of that year, there were 104 molecules in clinical trials or on the market that were based on companies’ original HTS campaigns (see chart, page 16). While not a huge number, it is an increase over 2004, when 74 molecules in clinical trials were based on original HTS campaigns. The top company in the 2005 survey said it had 12 compounds in the clinic based on original screenings done in 1994.
Wyeth did not begin conducting HTS in earnest company-wide until the late ’90s, Pasley says. “If I look at our entire pipeline all the way from very, very early in discovery—things that we’ve just finished up to right before they go to the clinic—we have impacted 50% of the programs that are sitting there.”
Drugs on the market
Bristol-Myers Squibb, Princeton, N.J., can point to one drug that they have on the market that emerged from an HTS program. Sprycel (dasatinib), an oral inhibitor of multiple tyrosine kinases, was granted accelerated approval in June for the treatment of adults in all phases of chronic myeloid leukemia with resistance or intolerance to prior therapy, including Gleevec (imatinib mesylate).
“It’s a good example of the serendipity of HTS because it was screened against a different target, but its activity was seen to be relevant to this target. Everyone here agrees
![]() |
that the molecule would not have been found to have that property if we had not stored the molecule in our storage system and then screened it in our laboratory screening group,” says Alastair Binnie, PhD, the company’s executive director of discovery informatics and automation. The patented molecule is not the screening hit, as there was a great deal of medicinal chemistry involved, but Binnie says there is “a clear genealogy, clear evolution from the screening hit.”
Martyn Banks, PhD, the company’s group director for lead discovery, profiling, and compound management, says the original screening hit was actually made in 1974; it then became part of the compound collection and, hence, the screening deck. When it was screened in 1997 in an HTS group for a tyrosine kinase, researchers realized that it might have advantages in a different tyrosine kinase and it was sent to the oncology group.
Binnie says the evolution of the molecule that eventually became Sprycel is typical, an excellent example of a promising screening hit being tweaked by a medicinal chemistry program, becoming a patented molecule, and making it to market. “There’s a clear path there, but even then, the history is actually quite complex. . . . It’s difficult to say that a molecule is a direct result of HTS, because in fact it’s the direct result of HTS and a great deal of other efforts as well.” Binnie’s comments underscore a fundamental factor at play in the backlash against HTS in some quarters over the last few years: A basic lack of understanding outside of, and occasionally in, the industry, as to the complexity of the discovery and development process.
Hype and reality
Bruce Koch, PhD, associate director and head of the HTS group at Roche Palo Alto, says that in the early days, one of the visions for HTS was that it could be used with combinatorial chemistry to produce enormous compound libraries. “HTS would then select compounds from the libraries and they would only need minor tweaking to become
clinical candidates,” says Koch, hastening to add that this was not the vision at Roche. “If that vision is the promise of HTS, then no, I don’t think fulfilling that is even possible. I think an alternate vision of HTS, and it’s one that I have and I think it’s shared by many of my colleagues, is that it’s a way of identifying weak starting points that medicinal chemists can then evolve.”
Banks says his first experience with HTS was in the mid ’80s doing natural product screening; combinatorial chemistry then emerged in the early ‘90s and seemed to get linked directly with HTS. “When we say HTS has not delivered on its promise, it was never clear to me what that promise was. Was the promise more drugs? If so, I don’t think HTS could ever achieve that goal because it finds hits that become the basis of chemistry programs, which, therefore, give rise to more sophisticated programs. But for some bizarre reason, which I never quite understood, this idea that HTS with combinatorial chemistry was going to deliver drugs was the mantra of the time,” he says.
“There was a very seductive dream that was sold,” Binnie adds, “that you’d be able to make so many compounds and screen them so quickly that you would basically short circuit years of medicinal chemistry and pharmacology. In hindsight, that just seems naïve that people would have believed that hype, but I think what happened to a certain extent was that there was a hangover, a reaction, when it became clear that it wasn’t that simple.” While no one would deny that HTS and automated compound storage and retrieval are essential tools in modern drug discovery, he says, the reality is that they are part of a long, complex process with many elements.
So while the technology has not delivered on the over-stated hype, it does allow researchers to perform efficient and consistent compound storage and screening. “There’s a very demonstrable value proposition in there, particularly today now that the technology is quite mature,” Binnie says.
Maturing technology
The maturity of HTS is a common theme amongst those who work in the field, and they offer a number of explanations as to why it evolved as it did. Wyeth’s Pasley says that in the 1980s, many of the screening operations launched came out of natural product screening. “Every company was looking for antibiotics, and they had put in place a rudimentary ability to look at tens of thousands of bacterial cultures,” he says. As company libraries began to get larger and larger, many, including Pfizer, Upjohn, and GlaxoSmithKline, began taking what they had learned and applying it to small molecules.
In the early 1990s, combinatorial chemistry arrived on the scene, allowing access to a large number of chemicals for screening, even for companies without many legacy compounds. At the same time, computer power increased dramatically, allowing for quicker data processing. These factors resulted in a number of instrumentation and reagent vendors introducing automated platforms, devices, and reagents. But it took researchers a great deal of trial and error to determine how to best use these tools.
“I think a lot of us floundered at various places, rapidly buying stuff and trying things that weren’t quite robust enough to generate high-quality data,” Pasley says. “A lot of people cut their teeth in the early ’90s at the bigger companies, and in the late ’90s we had groups of people who had now been doing this anywhere from five to ten years. They had come out of the trenches and knew what the downside for some approaches was, and I think it started to improve over the last seven to ten years. I think the investment in people, data management, and robust equipment has really started to help.”
The maturation of the technology and the continued investment should increasingly make an impact at other companies just as they have at Abbott, where several HTS leads have moved into clinical trials, says Burns. When he talks to colleagues at meetings such as the Society for Biomolecular Sciences, most have stories of hits they have identified using HTS that have moved into phase I, II, and III clinical trials.
Binnie says another indicator of the success of HTS is that it is bleeding over into other areas of research that were previously performed manually. “A typical organizational group that used to just do HTS will also do liability profiling, it will do activity profiling, it will do cancer screening, using the same basic technology platforms but a much broader portfolio of screening techniques,” he says. “We’re using the technology in context, really much more specialized than the full-on, full-deck, several-million-compound screen. We still do that, but we use the same basic platform for lots of other types of screening as well.”
Newer screening techniques
Researchers also continue to develop other types of screens. Roche, for example, has what it calls “iterative screening,” Binnie says. This technique entails screening a small proportion of the collection, then using the information gathered about the activity, the structural profile of the hits, to select another subset of the deck where it is believed the hits will be clustered. The information from that subset is used to screen a different subset of the deck. Thus, researchers work their way through the deck without necessarily screening every compound in the deck by using statistical predictive techniques based on the initial set of hits and then subsequent iterations of hits.
Burns says Abbott is using “affinity screening,” a technique they have used for years but which has recently become more mainstream as the company seeks ways to apply it and
![]() |
improve its performance. Affinity screening uses a variety of tools to measure the binding of a small molecule to a biological macromolecular target. Unlike activity-based HTS, affinity-based screening does not require knowledge of the structure or function of the target or target-specific assay development. Thus, compounds that modulate activity can be identified simply by binding to the target.
Roche’s Koch says another approach that is being used more commonly is fragment screening, a technique that is especially attractive due to the vastness of the chemical space. “The number of possible fragments is much, much smaller, so it’s a more tractable numbers problem,” he says. But, Koch adds, it has really only worked so far on targets in which, after fragments have been identified, they can be co-crystallized with the target proteins. “You can get some detailed information on fragment-target interactions. Where it’s worked, it’s been very powerful.”
A newer approach that many researchers are using is to look at more focused libraries, mostly of kinases and GPCRs, and less frequently, ion channels. Screening sub-libraries where the members are selected computationally, empirically, or both, to interact with a particular class of targets is an extremely effective and cost-efficient way of identifying bioactive chemical matter, says Amgen’s Tagari. “Our own protein kinase inhibitor library almost invariably reveals potent compounds that can be used to further investigate the biology and biochemistry of the target kinase, most often in a cellular setting. However, the very nature of these libraries results in both promiscuity, in terms of biological activity, and degeneracy, in the sense of overall chemical diversity and chemical intellectual property. So the utility of hits from such libraries as a starting place for a lead optimization program needs to be evaluated on a case-by-case basis.”
The same chemical matter?
Burns says one of the disadvantages of this approach is that if companies are simply creating the same chemical matter as other companies, it’s not going to help them have a focused library. He says Abbott and other companies are looking at things like non-traditional kinase cores and putting them together and creating libraries that might add significant intellectual property advantages if they get hits from HTS. “If you design the libraries to take advantage of cores that are not in the patent literature, for example, or in the public domain, that might allow you to find alternate chemical matter that would give you better IP position,” he says.
Pasley says large companies are running their corporate collections and investigating more targeted libraries, while there are some smaller companies focusing exclusively on a targeted approach. One is Astex Therapeutics Ltd., Cambridge, UK, which does fragment-based screening. The company features a technology called Pyramid, which uses high-throughput X-ray crystallography to define the binding of low affinity drug fragments to therapeutic target proteins. The fragments are then optimized into selective drug candidates that bind to these targets.
Vertex Pharmaceuticals Inc., Cambridge, Mass., has a large part of its library built around kinases and is developing a new class of oral therapies for the treatment of rheumatoid arthritis (RA). One of these is VX-702, a once daily, oral p38 MAP kinase inhibitor for RA that targets the production of specific cytokines. Vertex announced positive results from a phase II monotherapy study earlier this year and says it will initiate a three-month phase II study in patients with RA on a background of methotrexate. And Icagen Inc., Research Triangle Park, N.C., is focused on developing small molecule drugs that modulate ion channel targets. Its most advanced program is for ICA-17043, a compound for sickle cell disease, which they are preparing for a phase III clinical trial in collaboration with McNeil Consumer & Specialty Pharmaceuticals, a Johnson & Johnson company.
As for the future, Amgen’s Tagari says there are newer technologies that could further complement HTS. He says there have been significant advances in tools such as nuclear magnetic resonance and mass spectrometry that are directly applicable to fragment-based screening—either in mixtures, single fragments, or arrays—which show significant promise for the development of leads that affect protein-protein interactions.
“The general applicability will depend significantly on the chemical diversity of available fragments and the chemical feasibility of coupling fragments in a synergistic way,” he says. “These technical advances are, however, also relevant to more traditional screening and will provide important access to identifying allosteric inhibitors which might prove more selective than traditional active-site/binding-site ligands.”
This article was published in Drug Discovery & Development magazine: Vol. 9, No. 9, September, 2007, pp. 14-20.
Filed Under: Drug Discovery