Automated systems have pushed high-throughput screening bottlenecks further downstream, but some pressure points remain.
Catherine Shaffer
Shaffer is a freelance writer based in Ann Arbor, Mich.
Most major pharmaceutical companies now maintain a million or more chemical compounds for their high-throughput screening programs. As these gigantic libraries were being established in the late 1990s, it became apparent that old, manual methods of compound management were wholly inadequate to meet the demands of the screening systems. Bottlenecks developed at key points, such as weighing, plating, picking samples, and liquid handling as compound management groups struggled to supply scientists with compounds.
Most major pharmaceutical companies have now alleviated these bottlenecks by replacing crude, manual sample handling with robotics and automation. Compound libraries exist as liquid stock solutions within an automated compound storage facility. Automated compound management systems are challenged to provide thousands upon thousands of compounds in 96- 384- or 1536-well formats for several scientists, as well as quickly and accurately picking individual compounds from the library for custom sets.
For the most part, these systems have pushed the bottlenecks in high-throughput screening downstream to areas such as assay design and validation. However, some pressure points remain within automated compound management that will resurface as true bottlenecks in the future when assay design and validation catch up.
Balancing Speed, Accuracy
Speed and accuracy don’t mix well in the sciences. Rapid results all too often are hasty, sloppy, or inaccurate, so the speed of a compound management system must be balanced by accuracy. Human handling results in the greatest number of errors by far, so most compound handling systems strive to minimize human handling. Automated compound storage and retrieval systems generally rely on racks, carousels, and robotic arms to store and move compounds. The preferred storage method is in dimethyl sulfoxide (DMSO) solution at -20°C. The accuracy and precision of high-throughput screening results depend directly and crucially on the ability of the system to keep the compounds stable, prevent water ingress and compound degradation, accurately track compound identification, and, at a more basic level, run consistently without breaking down.
Since most storage and handling of compounds occur in liquid solution, one of the most critical points for potential sources of error is liquid handling. LabCyte Inc., Sunnyvale, Calif., has developed a novel solution to the challenge of moving liquids. Their “touchless” liquid handler, the Echo 550, uses focused acoustic energy (ultrasound) to eject a droplet from the surface of a liquid. Because the Echo 550 is touchless, there are no dispenser tips or other wastes generated. According to Timothy Spicer, PhD, a research scientist at Bristol-Myers Squibb (BMS), the Echo 550 is a very effective liquid handling tool. “The Echo 550 supersedes other dispensers in our hands. Nothing comes close in precision and accuracy, and it eliminates consumables in the environmental waste stream.”
Other solutions for liquid handling include piezo-based or pin-tool-type array dispensers. Also available are touchless solenoid-based liquid handlers. BMS uses a compound storage system from The Automation Partnership, Royston, UK, and has migrated some of its lead discovery activities closer to the central compound management system. “We directly support therapeutic areas by providing bioassay results to directly support medical chemistry programs,” says Martyn Banks, PhD, group director for lead discovery and profiling. “What was traditionally a therapeutic area role has moved into the lead discovery area, freeing up biologists to focus on higher value research. We’re being asked to directly support the discovery of the leads now. What’s also important to know is we’ll do secondary in vitro activity assays.”
By carrying out lead discovery assays centrally, rather than distributing them to therapeutic areas, BMS achieves a smoother integration of compound management and high-throughput screening, a result that eliminates speed bumps in the process. “One of the big issues that is left to be tackled is quantifying the amount of a compound in the assay,” says Banks. “What is the true compound concentration in the biological assay? We need to understand the solubility limits of the compound in the context of the assay in which it’s being run.” The Echo 550 provides an advantage in tackling this issue because it can audit the water concentration in a compound plate based on the known speed of sound waves through water and DMSO.
Water Risks
Water uptake is a major area of concern in compound management. DMSO is highly hygroscopic, so it’s nearly impossible to prevent water from entering the solution. However, the presence of water compromises the integrity of the compound, and over time the compound can degrade even as it is being stored. Every access event introduces the risk of water ingress: thawing, decapping, pipetting, etc. Available solutions for this problem include elaborate compound management facilities where compounds are stored and handled in a total nitrogen atmosphere.
However, TTP Labtech Ltd., Cambridge, UK, the originator of the haystack-type storage system, has developed a more workable system. The ComPound storage system is modular, barely larger than a laboratory freezer, and can store up to 200,000 vials. Modules can be connected to accommodate greater numbers of compounds. Instead of robot arms for retrieval, pneumatic tubes are used. The ComPound system can move unlimited numbers of tubes in parallel from multiple locations at high speed. TTP LabTech solves the problem of water ingress in the system by coupling it to the company’s Compiler Liquid Handling System. Compiler fills each tube with a plug of argon immediately upon decapping to minimize exposure to outside atmosphere.
Everything is done on a tube-sized scale, so the cost and logistics of the system are manageable. Openings in the cold storage area through which the tubes are sucked out are barely larger than a 1.4 mL tube and small volumes of argon are used to exclude atmosphere. The issue of water uptake has received a great deal of attention in the industry, but there is very little data available with real samples, says Jas Sanghera, PhD, commercial director of TTP LabTech. “The only way to find out how effective [a system is] is by actually measuring the degradation and water uptake. Wyeth has done an extensive test and so have some of our other customers. The amount of water uptake is very, very small to the point that it’s unmeasurable.”
Although human error is virtually eliminated with automated compound handling, mechanical error is still an issue. When processing hundreds of thousands or millions of samples, it is possible that the wrong tube could be delivered or that the wrong barcode could be attached to a sample. An error rate of even 0.1% multiplied by one million compounds would be 10,000 errors, and 10,000 potential hits missed. Strategies for controlling machine error focus mainly on redundancy in the system. If the wrong tube is selected, the error should be discovered when the bar code is scanned. However, formal studies of machine error rates are generally not available from laboratory equipment manufacturers.
A Future for Automated Compound Weighing and Management
Weighing of dry compounds for high-throughput screening is such a monumental chore that the dominant trend in compound management is the conversion of dry compound stocks to liquid stocks solvated in dimethyl sulfoxide (DMSO). The use of liquid stocks reduces the need for weighing of compound to only those times when the liquid stock runs out. However, the use of DMSO introduces new pitfalls relating to compound stability through freeze/thaw cycles and water ingress. Developers of laboratory automation technology continue to study the problem of automated compound weighing.
Innovate Engineering and Design in San Diego recently filed a patent application for a novel automated compound weighing system called the Nova Compound Collection System. It uses a disposable pin for collecting dry compound from a vial. The computer detects the weight of the compound on the pin and will repeat the collection until the weight is correct. It then transfers the compound into the destination vial. Decapping, capping, barcode scanning, and physical transport of the vials is included in the weighing process, which takes one to two minutes. The system is currently limited to non-metallic compounds with low moisture content and crystals of less than 2 mg mass. Cross contamination is a non-issue because of the use of disposable pins for handling the compound.
Other dry compound weighing systems are available using methods such as vibration, vacuum, valves, and Archimedes screws. The main problem with any automated compound weighing system is the fact that drug candidate compounds frequently have unusual physical properties that cannot be accommodated by the automated system. Until computers and robotics are smart enough and sensitive enough to handle these compounds, at least some manual weighing is necessary. One tool that can bring the manual weighing process as close as possible to the accuracy and efficiency of an automated system is the MossWeigh system that comes with the Moss 2002 Automated Assay Plate Preparation System from Zinsser Analytic GmbH, Frankfurt, Germany. The MossWeigh software guides the operator through a weighing process expedited by automatic doors, rapid taring, and automated bar code scanning and data storage, so the only task that remains is the actual handling of the compound.
Nuts and Bolts
Many potential bottlenecks in compound management, especially with increasingly automated systems, have to do with practicality. Often, the theoretical throughput of a system is greatly decreased by unavoidable, real-world problems, and the reliability of the automated system can become a potential pitfall. Sanghera says the main obstacles to maximum performance of compound management systems have to do with mechanical breakdown. “There’s no point in having automation if you’re going to spend most of your time repairing it . . . I’ve seen many bits of kit sitting in labs where people just want to forget that they ever bought them. Usually it’s because of reliability, rather than because when they bought it, it didn’t meet a particular specification.”
Another hurdle in compound management has to do with materials and consumables. Colette DeChard, manager of biological support for compound management at Merck, describes a state-of-the-art, fully-automated, compound management facility housed in a dedicated building in Rahway, N.J., that supports all research locations worldwide. This climate-controlled laboratory warehouse holds racks of carousels and cold boxes up to a 24-foot-high ceiling. Tray transfer robots move between shelves 100 inches wide and six feet deep, hefting racks of 96 vials or aluminum trays of plates. “Before, the bottleneck was manually picking the sample. Now we have robots that can pick 192 vessels in fifteen minutes. A human with semi-automated shelving would take an hour or two to pick those same 192 samples,” DeChard says.
Sixty CPUs control the robotics and databases, outnumbering human operators three to one. Merck’s system supports all nine worldwide research facilities, where tens or hundreds of thousands of compounds may be screened in a single day. This massive facility can pick and plate 12,000 compounds a day and replicate the same 12,000 compounds the next day. Such a massive, futuristic operation may seem unstoppable, but one of their greatest bottlenecks is keeping a supply of consumables such as plates, caps, and solvents, says DeChard. Run out of any of these, and the entire operation grinds to a halt. To avoid this, Merck uses software to manage supplies, maintaining a six-week lead time on all orders.
Weighing Pains
While investing in automation has paid off for Merck, one of the most persistent bottlenecks in compound management is also the one most resistant to automation. Compound weighing is a task that must still be done by human hands. No automated weighing system has proven to be capable of handling the variety of compounds used in pharmaceutical research. Many compounds are not simple powders, but sticky, gummy substances that stick to the sides of vials and to spatulas. When Merck first implemented its fully automated compound management system, it recruited every chemist and biologist in the entire company, all the way up to vice presidents, to help solvate the last 87,000 compounds in the collection. Over the course of 41 days, each scientist contributed 12 hours to weighing and solvating compounds. And although that initial hurdle has been overcome, Merck, like all other pharmaceutical companies, faces the task of repeating the weighing process for its entire library at intervals potentially as short as a few years in order to keep liquid stocks available.
One frustrating aspect of working with automated sample-handling systems is that the plates used for preparing or replicating samples might be incompatible with the intended automated high-throughput screening systems due to subtle differences in size, shape, or materials used in the plates. The mid-Atlantic chapter of the Laboratory Robotics Interest Group met in January to discuss a report from Drug and Market Development, a market analysis firm in Westborough, Mass. The report stated that “it may no longer be sufficient to provide increased throughput for screening while doing nothing to affect downstream bottlenecks in later-stage screening. Alternatively, it may no longer be sufficient to provide high-throughput screening solutions that fail to effectively interface with compound storage and retrieval systems.”
To that end, the Society for Biomolecular Screening formed a committee to recommend standard dimensions for microtiter plates. The Microplate Standards Development Committee recommends, develops, and maintains standards to facilitate automated processing of microplates on behalf of and for acceptance by the American National Standards Institute. Their recommended ideal microplate dimensions are 127.76 mm ±0.25 mm (5.0299 inches ±0.0098 inches) by 85.48 mm ±0.25 mm (3.3654 inches ±0.0098 inches) with a continuous and uninterrupted footprint all the way around the base. Standards for microplate height, bottom outside flange, and well positions have also been issued, and standards for side wall rigidity are in development. The adoption of a universal standard for microplates will facilitate the smoothest possible transition from compound management to high-throughput assay.
Seamless Integration
Yet another practical challenge to overcome is the interface between compound management systems and screening systems. Ideally, every compound order would be filled in a plate format that would mesh seamlessly with the screening apparatus and be labeled with a barcode recognized by a company-wide compound management informatics system. Then it would be delivered immediately to the proper location. Merck’s shipping time averages five days, and the company relies on individual compound-management groups at its research sites to do plate replication for individual screens.
An increasingly popular way of getting around interruptions in the flow of compounds between compound management and high throughput screening is combining the two functions, much like Bristol-Myers Squibb. This allows companies to maintain a close relationship between compound management and lead discovery; compatibility and interchangeability of plates; consistent labeling and compound tracking; short transit times for stability and integrity of compounds; and a minimum of liquid-handling steps.
The Michigan High Throughput Screening Center in Kalamazoo, Mich., is a nonprofit facility that seeks to support academic researchers, small companies, or even large companies who want to outsource. Scheduled to open in September, it will maintain a relatively small compound library based on a unique algorithm used to identify a diverse library of compounds. When it is up and running, the center’s compound library will combine TekCel storage units with robotic liquid handlers and the client reagents, enzymes, and other assay components. Ron Kilkuskie, PhD, the center’s senior director, says that although they plan to automate their compound management systems, they will proceed with caution. “We’ve talked to people about integrating the liquid handling robots with our TekCel system. Until we see it in operation we’re going to do it manually. We need to see how it works in our hands before we invest in software.”
Wherever compound management systems have been automated, they have surpassed other screening stages such as assay development and design in speed, and thus have not been used to their full potential. If future increases in speed are desired, it may be necessary to increase manpower or add second and third shifts so that computers and robots are not left idle while human operators rest. In the not-so-distant past, it was common practice for compound management groups to hand pick compounds from boxes on cold room shelves and weigh, dissolve, and plate compounds for assays, a slow and error-prone process. The advent of automation has accelerated compound management from the slowest stage of drug discovery to the fastest.
Filed Under: Drug Discovery
