Surrendering to the robot army: why we resist automation in drug discovery and development

As scientists, the prospect of acquiring and learning how to utilize new laboratory instrumentation, such as mass spectrometers, generally fills us with enthusiasm if not glee. Some scientists are continually striving to obtain cutting- or bleeding-edge technologies, but a surprising number are slow to adapt to new tools and understand how best to utilize them within our daily work. Although LC–MS almost completely superseded liquid chromatography with UV or fluorescence detection at a very rapid pace, the adoption of complementary liquid-handling robotic tools to prepare samples for analysis by LC–MS has not been as universal or as rapid. In this editorial, I will compare and contrast the reasons for and against utilizing automation. Hopefully my justifications for automation will be so compelling that I convince the unbelievers to rethink their resistance to automation, and, thus, surrender to the robot army.

Indoctrination into the robot army should begin with the litany of advantages that automation provides. Automation has been leveraged for a broad range of drug discovery assays including primary enzyme activity screens (identification of ‘hits’ against a specific protein target), microscale combinatorial-based synthesis and preparation, in vitro absorption, distribution, metabolism, excretion and toxicology screening and preparation of biological matrices from animal and human studies.

Robotics are an ideal fit for repetitive tasks, aliquotting large numbers of samples, preparing standards, blanks and quality control samples in large volumes and multiple replicates. Consistent precision and accuracy are provided, without concern for human error. To be fair, automation can be consistently wrong, if not correctly programmed.

Depending on the instrument’s capabilities, smaller (nl) volumes can be handled by automation rather than manually. Ergonomic issues, such as carpal tunnel syndrome, from manual pipetting for extended periods of time can be avoided. Human exposure to biohazardous or toxic materials can be reduced, although the robot itself may become contaminated. Automation drives a significant level of standardization across a large group of scientists or experimental types, as the inputs to the liquid-handling program must be correctly entered and consistently executed. Although this requires extensive discussion initially to reach agreement on the actual experimental task execution, once decisions have been made, the automated method will be done consistently without any need for re-training. Perhaps the most attractive benefit of automation is that robotic conduct of mundane, repetitive tasks, frees up scientists’ time to focus on more difficult but rewarding challenges.

Nevertheless, a number of organizations have yet to embrace automation. The obstacles to this adoption can be categorized as technical, process-related or psychological

Technical barriers to automation

As many scientists are inherently frugal by nature, the most immediate stumbling block to automation is the price tag. Although basic 96-well pipetting workstations may sell for <US$50,000, more sophisticated platforms that include both single channel and 96- or 384-well parallel liquid-handling functionalities represent at least a six-figure capital investment. Deckware components and consumables, such as pipette tips, are precision crafted and generally more expensive than those used for manually pipetting. For larger organizations and capital budgets, the initial investment is generally doubled in the interest of redundant systems to minimize disruptions to a productive workflow. If one system malfunctions, work can then be shifted to the back-up system without downtime. The equipment will need to be maintained, and some time will be needed initially to calibrate and validate the robotic system’s capabilities, and establish regular calibration or instrument verification procedures to ensure it functions as expected.

Although the hardware within a liquid-handling system is relatively robust, over the course of the instrument’s lifetime, improvements will undoubtedly be made available for both hardware and software components. In many cases upgrades can be performed efficiently, but in others this may be cost prohibitive or incredibly labor intensive, such as cutting a hole in the middle of the deck and thus fundamentally shaking the instrument on its foundation. Software validation and version control can be particularly problematic in a regulated environment, that is, when working according to GLP [1].

A key consideration and sometimes a very significant barrier to automation is the ease of integration of the automation with other laboratory hardware and software tools. Many electronic laboratory notebooks, laboratory information management systems or LC/MS software packages may not be easily or seamlessly integrated with automation outputs. If this integration is critical to productivity and successful execution of experiments, utilizing automation may exceed the grasp of many. However, numerous laboratories utilize sample preparation automation as a standalone unit or ‘island’ to circumvent this very problem.

Finally, the organization will need to invest in training their scientists to program and operate the automation. Once trained, the scientists should be considered higher flight risk (i.e., to leave the company), as they now possess highly marketable skills. Talent retention of these scientists thus becomes more important for management.

Process-related barriers to automation

The way in which automation is utilized; that is, its process, plays a critical role in successful implementation. Multiple strategies can be leveraged, but the organizational structure and workflow should be considered when designing a process. Generally, automation is used via either an open-access or centralized model. An open-access process is essentially the democratic usage of robotics. Usually a highly trained scientist or ‘super user’ oversees the instrument installation and verification procedures. This super user then creates and maintains a menu of programs for other scientists to use for their experiments. A daily or weekly schedule may be utilized, where users select time windows according to their needs and workflows. Open-access approaches are popular with small- to medium- size groups where daily scheduling adjustments can be smoothly handled in a collaborative manner. Users benefit from the training required to operate the system, and gain proficiency and confidence in their new skill sets (and resumé enhancement). One disadvantage of an open-access process is that access to the robotics may become a bottleneck to completing work. High-demand time windows, such as mid-morning or afternoon, may create user delays as they wait for other scientists to finish their tasks. Assigning responsibility for ordering supplies and maintaining the system is another important consideration. When multiple users may assume that someone else is accountable, the group may run out of supplies or the instrument may malfunction. In other words, ‘when everyone is accountable no one is accountable’ [2].

For large groups of individuals (>10), utilizing a single point of contact or operator model is an attractive alternative to open access. In this process, one scientist is responsible for the robotic assays or tasks submitted to the work group. This individual oversees the execution of multiple batches over the course of a day, maintains the system and delivers the finished product to other scientists. Accountability is clear and operational efficiency is maximized. This task is not particularly exciting, and thus may result in boredom and decreased job satisfaction. In my group at Merck, we have implemented a rotation system that allows scientists to be trained on robotics usage and rotate through responsibilities for a short period of time (2–3 months).

Process implementation that includes automation can be particularly challenging in a regulated environment, as mentioned earlier. Once methods have been validated using certain equipment or procedures, modifications while the work is ongoing can be quite difficult. In a clinical environment where analytical work can span years or decades, this becomes even more serious. Thankfully, the drug-discovery environment is more amenable to change, as evidence by the proportionally greater usage of automation in screening experiments.

Psychological barriers to automation

Technical or process barriers notwithstanding, the psychological barriers to automation can halt the advancement of the robot army (implementation of automation) in its tracks. The most serious threat is job security – the robot will replace me. Maslow’s seminal work in the psychology of human need places self-actualization (creativity, problem solving) and esteem (achievement, respect from others) as the top two elements of happiness [3]. Despite a laundry list of examples where technology has significantly increased our quality of life – the dishwasher, washing machine, telephone, automobile, mass spectrometer, autosampler – scientists who have spent years conducting a certain experiment manually, will immediately respond with resentment and suspicion to automation. Ideally, implementing automation will not result in immediate job loss. Instead, automation should free scientists from the mundane to focus on more challenging aspects. Perversely, some will still resist because they enjoy the mundane.

The complexity of specific automation tasks may lead to difficult relinquishing of control and trusting the robotic end-product. Some individuals also grieve the loss of their sense of accomplishment on successful manual completion of a task. Technical skills are often the result of years challenging oneself to pipette more and more samples and finish more complex studies in shorter periods of time. When manual pipetting is replaced with automation, people may struggle with the loss of pride in their own handiwork.

Another insidious psychological barrier is the ‘not invented here syndrome’. Despite any level of altruistic intention to leverage the collective knowledge of the scientific community inside or outside the organization, many resist automation because they were not involved in its development, either technical or process. Commissioning a core team early during the implementation process is important, but as usage is widened across a group, as many people as possible should be engaged to refine and continually improve robotic assays and usage. This creates a sense of ownership and commitment to automation.

Regardless of which psychological barriers are prevalent, automation team members will need a healthy dose of patience and persistence to succeed. Another key ingredient is management commitment to utilize the automation, rather than simply installing another expensive black box that collects dust in a corner of the laboratory. Generals in the robot army should brace themselves with a healthy dose of intestinal fortitude, but the long-term benefits will far outweigh the pain.

Looking forward, the ultimate benchmark of automation, or victory cry, will be its utilization in a contract research organization (CRO). CROs are significantly more pragmatic than pharmaceutical or biotechnology companies, with a universal focus on the profit and loss equation. As their customers increasingly leverage automation and thus expect it when placing work at a CRO, the robot army will advance even further until all possible combatants are assimilated.