The Impact of Automation on Drug Discovery
William P. Janzen
Sphinx Pharmaceuticals
A Division of Eli Lilly and Company
4615 University Drive
Durham, NC 27707
http://www.netsci.org/Science/Screening/feature08.html
Automation n 1. in manufacturing, a system or method in which many of all of the processes of production, movement, and inspection of parts and materials are automatically performed or controlled by self operating machinery, electronic devices, etc.. 2. any system or method resembling this using self operating equipment, electronic devices, etc. to replace human beings in doing routine or repetitive work. 3. The condition of being automated.[1]
Automated systems have been used in drug discovery for many years. That they have had a major impact on this process is indisputable. However, an assessment of their impact is difficult for a number of reasons. Many of these systems are applied in areas, such as High Throughput Screening (HTS) and Combinatorial Chemistry (CC), which have incorporated automation since their inception. In addition, most of the tasks performed in these areas are highly proprietary, which at best delays disclosure of details and at worst prevents it altogether. Empirical productivity analysis is further hampered by a tendency to publish projected outputs (capacity) rather than actual figures. Sphinx Pharmaceuticals is a unique position in that the systems in HTS[2] have operated in both manual and automated modes, thereby lending a perspective on automation and productivity.
Before examining automation's uses, one must define some subtypes of automation. Many common laboratory devices could be classified as automation. However, let us make the distinction that equipment must perform a function autonomously to be considered automated. This excludes most lab equipment. The remaining devices, although they may require human initiation and intervention, are clearly "Automation". Automated equipment can be further divided into three primary categories; hand-held, unit automation, and integrated automation. Each of these types has valuable uses in the Discovery Laboratory.
As is implied by its name, hand-held automation includes devices which not only require human participation to function, but must be held and controlled. Examples of this type of equipment are the multitude of automated pipetters, single and multichannel. By facilitating the ability to easily dispense multiple aliquots and to add reagents in rows and columns of microplates, these devices enabled the use of microplate technologies in areas which cannot afford large scale automation or where its application is inappropriate. They are heavily used at Sphinx in both HTS[2] and CC[3] for development work and low throughput assays. Other devices typical of this category of automation are the multitude of hand-held and portable computing, scanning, and calculating devices. Hand-held bar-code scanners enable mobile sample tracking and inventory control with the option of maintaining radio frequency based links to lab computer systems, while the ubiquitous laptop computer is indispensable in an environment where equipment requires microprocessor control for diagnostics and operation.
While hand-held automation has found increasing use in all phases of discovery research, the most commonly used type of automation remains unit automation. Unit automation includes all those benchtop devices that have the capacity to conduct process unattended. This type of equipment makes up the subsystems of fully integrated automated systems and, admittedly, there is a very fuzzy distinction between the two types. Included here are most of the common readers and detection devices found in biological laboratories as well as workstation robotics. Workstations generally are defined as liquid handling X-Y-Z robots capable of aspirating and/or dispensing liquid. There are an increasing number of workstations that also incorporate basic plate moving capabilities and fall somewhere between unit and full automation. Workstations are also being designed to perform functions specific to Genomics and CC laboratories such as picking plaques, resin dispensing, and spotting TLC plates.
The picture that comes readily to most peoples' minds when discussing automation is one of fully automated systems. In these systems, unit automation is integrated with a device (most commonly a robotic arm) for transferring tubes or microplates. The complexity of these systems can range from simple plate feeders to large systems capable of performing complex HTS assays. Many equipment manufacturers have developed these systems with a variety of architectures. The introduction of robotic fingers for existing industrial arms that are capable of gripping and moving microplates made feasible the integration of modular unit automation. Many systems are now created by professional Integrators whose business is the creation and assembly of systems rather than the manufacture of equipment. A recent innovation has been the introduction of control software capable of scheduling tasks and conducting error recovery. With these packages it is possible to analyze and optimize a process off-line and then make appropriate changes to system hardware. Error recovery allows a system to continue operation even when subsystems or unit automation fails.
The simplest analysis of the impact of this technology would be to examine the relative outputs from each. As a model, let us consider a common HTS assays technique, Scintillation Proximity Assay (SPA; Amersham, plc). Two nearly identical screens measuring protein ligand binding were performed at Sphinx using SPA (table 1).
Table 1
Impact of Automation on an SPA Binding Screen
| Assay #1 | Assay #2 | Assay #1 or #2 | |
|---|---|---|---|
| Automation Employed | Unit Automation Tecan 5052 (figure 1) ICN Multidrop (figure 2) Wallac Microbeta (figure 3) |
Full Automation (figure 4) Hewlett Packard ORCA Arm Tecan Genesis SLT 96PW Scitec carousel Wallac Microbeta |
Hand-held Automation Matrix (figure 5) Wallac Microbeta |
| Samples Tested/Day (Maximum) | 17,760 | 10,080 | 960 |
| FTE's Required1 | 2 | 1 | 1 |
| Tested/FTE/Day | 8,880 | 10,080 | 960 |
| Protocol | Add buffer Add test compound Add ligand Add protein Shake Incubate 90 minutes Add SPA Beads Incubate 2-48 hours Count |
Add buffer Add test compound Add ligand Add protein Shake Incubate 90 minutes Add SPA Beads Shake Incubate 2 hours Count |
... |
1. Full time Employees required for all facets of the screen including reagent preparation, screen operation, and data reduction and review.
The results are counterintuitive. One would expect that a fully automated system would be much more efficient and operate at a higher rate, but the system employing unit automation was considerably faster and both operated at approximately equal efficiencies. A closer examination of the process explains this result. In Assay #1, the incubation timings are very flexible. This allows the operators to process microplates in a batch fashion. After rate limiting steps, plates can be held and processed as a large batch. Also, a fundamental difference between fully automated and semi-automated assays is the ability to parallel process. By assigning duties to two or more people, tasks can be conducted simultaneously. So why was Assay #2 automated? The answer lies, again, in the incubation times. The tolerances of the incubation times are much smaller. In fact, to conduct this assay, all counting had to be done the same day the assay was run. Eighteen hours of operation were required to reach the target minimum daily rate of 50 microplates per day. For reproducibility to remain at required levels, precise timings on all plates must be maintained throughout this period. Clearly, this assay would have been very difficult if run without electronic scheduling.
This result is typical of the efficiency analyses of targets done by Sphinx. In every case to date, throughput will be higher if semi automated processes with unit automation are employed, yet we own four fully automated systems. The justification for this is that although manual operations may be faster, there are sufficient cases like the one detailed above which require the use of full automation. It may be that a screen could be faster if unit automation is applied, but it has a repetitive task for which no unit with multiple plate capacity exists. This replacement of a repetitive task with an automated system is a reasonable use. In addition, safety becomes a concern. When handling large amounts of radioisotopes, infectious materials, or biological hazards, the reduction of exposure to laboratory personnel justifies the use of these systems.
However, the question remains open, what has been the impact of automation on drug discovery? At a 1992 at a HTS meeting, the highest reported throughput was 9,216 samples per day in three assays[3]. In 1995 at the Society for Biomolecular Screening meeting in Philadelphia, PA, it was common to discuss rates of 10,000 - 20,000 samples per day in multiple targets. This increase has been largely enabled by automation. Even though full automation may not allow faster testing, it will usually allow more testing per employee. Connected with the common practice of controlling hiring more closely than capital expenditures, automation may be the only solution for increasing throughput. So it could be said that automation is enabling by gaining management approval for expansion.
The HTS program at Sphinx has also followed this general trend. Relative throughputs, supporting personnel levels, and enabling technologies are shown in Table 2.
Table 2:
Enabling Technologies
| Average Daily Throughput | Personnel Level (FTE's) | Enabling Technology | |
|---|---|---|---|
| 1993 | 820 | 13 | Workstations with assays conducted on a single unit. |
| 1994 | 5,100 | 13 | Modular unit automation for screen operation. |
| 1995 | 30,000 | 21 | Automated compound preparation. One fully automated system in late 1995 |
| 1996 | 40,000 | 23 | Modular Unit Automation with stacking plate
feeders. Second fully automated system |
While staffing levels remained fairly constant, throughput increased greatly. As can be seen here, the largest impact was the introduction of modular unit automation. However, not included in this table is the fact that several of the targets tested during this period would have been significantly delayed or slowed without fully automated systems.
Another benefit of automation in the biological sciences that is often overlooked is the application of processes developed for HTS, or other areas widely employing automation, in research and methods development. By incorporating very basic hand- held or unit automation and some of the basic techniques such as the use of microplates and subsequent column and row wise reagent additions, significant reductions in development times and data output for research areas have been attained. Automation is generally less available in the research arenas for a number of reasons. Unless there is some connection with a facility employing these techniques, Researchers may not be aware of their power. It is also costly and may not be affordable to academic researchers. A more fundamental problem, is the required flexibility which must be available in a basic research facility.
Automation has its downsides as well. One must be very careful when automating a process, particularly a biological assay, that basic properties are not changed. Seemingly simple changes, such as the order in which reagents are added, can have a major impact on the relevance of results. Another consideration is that once a process is automated, it may be very difficult to change. In manufacturing it is common to establish an automated process and have it remain constant for years. This is rarely the case in Discovery Research. At Sphinx, the average duration of a HTS screen is 4 months. The time required to modify equipment between assays must be considered when making decisions on the level of automation to employ. Wherever possible, similar targets are used to replace existing targets on automated systems. This can lead to decisions driven by convenience rather than efficiency. Over-reliance on automation can be a problem. If laboratory personnel are used to having a machine which can perform routine tasks, they may, again, make decisions based on convenience rather that efficiency. Probably the most significant problem with automation is downtime. It is a fact of life in HTS that, even with routine maintenance schedules, complex equipment breaks down. Processes that are dedicated to automated systems must have this built into their projected timelines and contingency plans must exist for equipment failure. Finally, advanced data handling techniques must be in place to process the huge number streams flowing from these systems.
In conclusion, it appears that the best is yet to come. As automated systems are more heavily utilized in the Discovery Process, innovation is accelerating. With the introduction of smart scheduling software which can dynamically schedule processes to accommodate midstream changes or errors, we will see improvements in system flexibility and reliability. The ability to continue to run even after a peripheral fails will need to be an integral part of these programs. Modularity in automation is already increasing. The incorporation of stacking systems in unit automation makes it possible to feed a variety of devices, thereby allowing a single manual step to move batches of 20 or more plates. In fully automated systems the use modular components is already common. What is need is the ability to change them without system downtime to prevent work stoppages from upgrades and peripheral repair. And finally, a greater incorporation of these techniques into basic research from the Genomics effort and greater involvement of researchers in the HTS and CC processes can drive improved efficiencies in all facets of the biological sciences.
Acknowledgments
The author would like to acknowledge the contributions of all the members of Biomolecular Screening. In particular, Fred Hubbard, Frankie Mylott, Donna Stevens, Tome Gurganus, Mark Stamper, and Tom Stevens have contributed greatly to the work presented here.
References
- Webster's New World Dictionary, second College Edition, 1970, The World Publishing Company.
- Janzen, W. P.(In Press) Mammalian Cells in High Throughput Screening, Proceedings of the 14th European Society for Animal Cell Technology.
- Meyers, H.V. et al (1995) Molecular Diversity, 1, 13.
- Leichtfreid, F (1992) Novel Approaches to High Throughput Screening Automation, Data Management Technologies in Biological Screening, SRI International
Figure 1
Unit Automation: The Tecan 5052
Full Image 170 Kbytes (756 X
504 pixels)
Return to the paper
Figure 2
Unit Automation: The ICN Multidrop
Full Image 170 Kbytes (756 X
504 pixels)
Figure 3
Unit Automation: The Wallac Microbeta
Full Image 170 Kbytes (756 X
504 pixels)
Figure 4
Full Automation: Hewlett Packard ORCA Arm, Tecan Genesis
SLT 96PW, Scitec carousel, Wallac Microbeta
Full Image 170 Kbytes (756 X
504 pixels)
Figure 5
Hand-held Automation Matrix
Full Image 170 Kbytes (756 X
504 pixels)
NetSci, ISSN 1092-7360, is published by Network Science Corporation. Except where expressly stated, content at this site is copyright (© 1995 - 2010) by Network Science Corporation and is for your personal use only. No redistribution is allowed without written permission from Network Science Corporation. This web site is managed by:
- Network Science Corporation
- 4411 Connecticut Avenue NW, STE 514
- Washington, DC 20008
- Tel: (828) 817-9811
- E-mail: TheEditors@netsci.org
- Website Hosted by Total Choice