Billions of screens have produced … what?
Critics of high-throughput screening ask: Is the approach of building a bigger haystack really the best way to find more needles?
Nanoliter acoustic compound dispensing technology;custom-made benchtop enclosures with automated liquid handling vortexers andmixers; chemical libraries that have grown to include more than 1 million smallmolecules and grow by 20 percent every year; ministores that have the capacityto store million of compound in 384 well microplates at -20º C that can becherry-picked from the same or different collections—and that's just a brieflook at the advanced equipment side.
Elsewhere, there's induced fit docking;rational drug design; structure-based lead optimization; integration ofexperimental and in-silico data bycross-functional expert teams and dozens of other approaches.
It's all subsumed under the single rubric of high-throughputscreening (HTS), which is used in one form or another in hundreds, if notthousands, of labs around the globe.
HTS is defined asa method for scientific experimentation especially used in drugdiscovery and relevant to the fields of biologyand chemistry.
Using robotics,data processing and control software, liquid handling devices and sensitivedetectors, HTS allows a researcher to quickly conduct millions of biochemical, genetic orpharmacological tests.
Through this process, one can rapidly identify activecompounds, antibodies or genes which modulate a particular bimolecular pathway.
The results of these experiments provide starting points for drug design andfor understanding the interaction or role of a particular biochemical processin biology.
Yet after 25 years of HTS, we have little by way of NMEs thathave contributed to the cure of disease or alleviation of debilitatingsymptoms.
Several years ago, a respected U.K. researcher, Dr. David Horrobin,vented his frustration by decrying the process altogether. He noted thatestimates of the ratios of compounds synthesized to marketed drugs at the timeof peak success of Nobel Laureates Black, Bovet, Elion and Hitchings was about100:1; most of the industry from about 1960 to about 1990 saw about 10,000:1 to30,000:1; Big Pharma since the introduction of combinatorial chemistry and HTS,well over 1,000,000:1.
Horrobin asked the question, "Is the approach ofbuilding a bigger haystack really the best way to find more needles?
Dr. Stephan Heyse, head of Genedata's Screener businessunit, doesn't care for the haystack analogy because he says HTS has changed. Hethinks the "more needles" approach has been supplanted by a "sharper needles"goal.
"It's more like a well-tended field where you already know alot about what's in each row," he says. "We're evolving toward biology-richinformation that goes beyond simple endpoint assays and uses better detectiontechnologies, such as optical assays and ion channel readers, to make resultsmore trustworthy. The problem has always been that what you saw at the firstfilter will always be out there and can affect basic business decisions. Youcan't always afford to reproduce scans to get to the next level. High-contentscreening, for example, provides a much broader basis for decision-making. Youcan generate active plus toxicology information, for example. Maybe weakactives that have a good tox profile are more important than just strong hits."
Summarizing the current state-of-the-art technology, Heysenotes that classical primary screens continue to be performed in high throughput—i.e.,millions of wells. As new technologies such as high-content screening andtime-resolved fluorescence, label-free and electrophysiology methods delivermore information per well and screened compound, data management and analysisbecome more complex.
Mastering these challenges yields more precise informationat the HTS stage on compound mode-of-action and potential therapeutic window.Complete bioactivity profiles of compounds are compiled from sets ofhigh-throughput primary and secondary screens, enabling optimized decisions oncompound progression into the hit-to-lead phase.
At SchrödingerInc. screening can vary from ligand-based similarity searches where thousandsor tens of thousands of molecules are screened per second to much more refinedand specific studies such as induced fit docking, explains Dr. Woody Sherman,vice president of applications science.
In May 2010, his group reported theresults of a large-scale, ligand-based virtual screening study, with the goalof improving database enrichments.
The study involved 11pharmaceutically relevant targets to investigate the interrelation between 8two-dimensional fingerprinting methods, 13 atom-typing schemes, 13 bit scalingrules and 12 similarity metrics using the new cheminformatics package Canvas.
In total, 157,872 virtual screens were performed to assessthe ability of each combination of parameters to identify actives in a databasescreen. In general, fingerprint methods such as MOLPRINT2D, Radial andDendritic that encode information about the local environment beyond simplelinear paths outperformed other fingerprint methods. Atom-typing schemes withmore specific information, such as Daylight, Mol2 and Carhart were generallysuperior to more generic atom-typing schemes.
Enrichment factors across alltargets were improved considerably with the best settings, although no singleset of parameters performed optimally on all targets.
Kinases remain animportant drug target class within the pharmaceutical industry, Sherman notes, buthe adds that the rational design of kinase inhibitors is plagued by thecomplexity of gaining selectivity for a small number of proteins within afamily of more than 500 related enzymes. He and his team have developed acomputational screening method for identifying the location and thermodynamicproperties of water molecules within a protein binding site that can yieldinsight into previously inexplicable selectivity and structure-activityrelationships. Four kinase systems (Src family, Abl/c-Kit, Syk/ZAP-70, andCDK2/4) were investigated, and differences in predicted water moleculelocations and energetics were able to explain the experimentally observedbinding selectivity profiles. The successful predictions across the range ofkinases suggest that this screening methodology could be generally applicablefor predicting selectivity profiles in related targets.
"Understanding kinase selectivity is key to developingeffective therapies that don't have side effects," he concludes.
As Sherman's work reveals, screening that predicts or confirmsexperimental observations can answer, in silico,fundamental questions about molecular interactions and be used as an importantpart of the drug development process.