With the current rapid developments in techniques to process, analyze, and store large amounts of data also comes the opportunity to produce more data1,2. In all sectors within life sciences, this trend has increased the application of large-scale screenings3, replacing the old strategy of sequences of smaller analyses4,5. The number of samples in such screening is generally so large that humans cannot handle it anymore, and laboratory automation solutions have to take over. However, when not performed properly, laboratory automation can create a bigger problem than the one it was supposed to solve6. Therefore, the aim of this article is to explain relevant commonly used terms in the field of laboratory automation, provide factors to be considered before automating a laboratory, summarize requirements for a successfully automated laboratory, and identify typical laboratories and sectors that fulfill the requirements and, therefore, could be easily and successfully automated.
From manual tools to laboratory automation
Manual tools are generally designed to perform one particular task when properly operated by humans. There is usually a mechanical link between the manual action and the performed task7. Since the performance of the task highly depends on the skills of the operator8, machines and devices have been built to minimize inter-sample and inter-operator variability. These machines and devices are still designed to perform only one task, but the human factor is much smaller, since the human action is not directly coupled to the performed task (e.g. the operator presses a button and a component of the device makes a pre-defined movement)9. When multiple tasks have to be performed sequentially in a process, multiple machines or devices can be coupled into a partial laboratory automation solution6. This solution should also contain a system for movement of samples from one component to the other. However, human intervention is still required in certain steps of the process before and/or after the automated tasks. When an entire process – all tasks between sample input and data output, including a sample transport system – can be performed without human intervention, this is called total laboratory automation (TLA)6.
LAS, API, LIS, and LIMS
All components in a partial or total laboratory automation solution are controlled by the laboratory automation system (LAS). Communication between the LAS and all separate components is bidirectional rather than unidirectional10, in order for the LAS to receive feedback from the components regarding the status of the performed tasks (see: ‘Requirements for successful laboratory automation’). A LAS can be relatively simple when all components work with the same programming language and system, but can also become very complex when extra ‘layers’ have to be built in for ‘translating’ the programming languages from all separate components to the global LAS11. Developers of machines and devices can choose to make their application programming interface (API) open rather than closed. This means that users can get full or partial access to software applications that can communicate with the controlling software of the device, beyond the generally accessible user interface12. This open API facilitates case- and LAS-specific incorporation of the device into a larger automation solution. However, from a marketing and economic perspective, a closed API could be more beneficial in certain cases: the developer keeps full control over the device and controlling software, and can make a profit from designing and building complete solutions for partial or total laboratory automation13.
When a device has produced data, it sends it to the LAS, which in turn forwards it to the laboratory information system (LIS) or laboratory information management system (LIMS) (Figure 1). The difference between the LIS and LIMS is that the LIS is most commonly used in clinical laboratories, and focuses on the patient: patient’s demographics, clinical analysis results, and patient reports14. A LIMS can be used in all sorts of life sciences laboratories, and it focuses on the samples (e.g. experimental data, sample management, and quality control)14. Devices can also directly send data to the LIS or LIMS, but this requires confirmation to be sent back to the device that all data was received properly11.
Figure 1. A schematic representation of a total laboratory automation system with devices, LAS, and LIS/LIMS (adapted from Hawker, 2017)11.
Considerations before automating a laboratory
No two laboratories are the same, and an automation solution that works perfectly for one, may be completely unsuitable for the other. Therefore, a number of considerations need to be made to determine the optimal automation solution. Here, if the cons preponderate, partial or no automation might be the better choice for a laboratory.
Pros of total laboratory automation
Relieve employees from repetitive work15. Generally, with large-scale experiments comes the necessity to perform more repetitive work that might be extremely tiresome for researchers, as it has to be performed consistently and for long periods of time. However, repetitive work is one of the strengths of devices and automated systems16 and, therefore, increases consistency of conducted experiments15. This should also make experiments more reproducible, which is extra relevant considering the reproducibility crisis the scientific world is currently dealing with: 70% of scientists cannot reproduce the data of other scientists, and 50% cannot even reproduce their own data17. With automated systems taking over the repetitive work, human employees can spend their time on more conceptual work at their intelligence level, which should make the job more satisfactory. Also, the risks of employees being exposed to biological hazards or developing repetitive strain injuries (RSI) are much lower in automated laboratories6,15,16.
Increase productivity10. Work in automated laboratories is not restricted to office hours: it can be performed 24 hours per day, 7 days per week. This provides results in a shorter time frame, also considering that the lag time due to starting up experiments in the morning or after the weekend is absent15, so time is used more efficiently. Besides that, automated systems generally require smaller sample volumes6 – which improves cost-efficiency and/or allows for large-scale experiments – and can support visually or otherwise impaired laboratory members, promoting independency and increasing productivity per head.
Before significant improvements in productivity can be made, a transition period, in which the new technologies are introduced, should be allowed1. In this period, employees are getting used to the new systems, and systems sometimes still need minor tweaking or adjustments to function optimally. Therefore, errors may be made more often, reducing total productivity – possibly even compared to the pre-automation situation. This is a short-term investment for the long-term benefits but has to be overcome for automation to work best.
Improve sample management6. Humans tend to store objects via a self-invented system that seems to be convenient at the moment but turns out to be inconvenient for tracing back those objects6. Automated laboratory systems store samples and keep track of the storage location in the LIMS, where search functions can make it very straightforward to trace a sample. This is particularly relevant for the rapidly growing biobanks18, containing so many samples that makes finding a lost sample very difficult – if not impossible. Besides the sample location, the LIMS also facilitates structured data management and analysis, where humans often follow a similar strategy as for the sample storage.
Cons of total laboratory automation
Dependence on machinery6. Even the most advanced devices and systems can break down, and in an automated laboratory, a component breaking down without sufficient backup (see: ‘Requirements for successful laboratory automation’) can stagnate the entire process. If researchers have become less skilled in that particular task due to a lack of practice6, the dependence on extremely reliable machinery is very large.
Resources6. Automating a laboratory requires significant financial short-term investments. Even though these investments generally pay off in the long run, the budget has to be available before starting the automation. Besides that, specialized and often expensive supplies may be required for the system to function4,6, and these investments may be too costly for certain laboratories. Partial or total laboratory automation solutions can be very sizeable systems1, so the space to position them has to be available. In newly built laboratories, the required space can be taken into account during the design phase, but existing buildings can provide large restrictions in space, but also in the positioning of electricity, water, and gas supplies1. These restrictions can lead to either physical or technical bottlenecks. Since the speed and efficiency of a process are highly dependent on the slowest and least efficient step19, these bottlenecks can nullify all gained time and efficiency of the automation solution.
Standardization is not always possible6. When working with samples that require individual treatments without numerous repetitive steps, laboratory automation can be redundant – if not inconvenient. Even within an automated system, it has to be considered whether the possibilities for special requirements like sample prioritization, repeated testing, or dilution testing need to be included in the system, or these samples should get individual and separate treatments4.
Requirements for successful laboratory automation
While the emphasis of developments in automation used to be on the hardware, the largest progress is currently being made in the field of process control software (LAS, Open API, LIS, LIMS) for smooth integration of components5. In order for laboratory automation to become successful, the following requirements for the combination of software and hardware should at least be met.
Processes are optimized before the integration of an automated system4,20. There is much more potential gain – with lower investments – in optimizing a process than in automating it. Therefore, before considering automation, the optimization should be checked. There are numerous systems to do this, like Lean (for streamlining the work), Six Sigma (for improving quality), and so on20.
There is communication between all components10. As indicated in the definition of the LAS, there should not be unidirectional control over the devices, but bidirectional communication between the LAS, LIS, or LIMS, and all devices and components. In this way, every step in the process is checked via confirmations of the performed tasks.
Automation reaches all steps of a process21. Steps of a process that are not included in the automation solution will become the bottlenecks for the entire process. This does not mean that all steps should by definition be automated21, but they should at least be taken into consideration when designing the entire solution. These steps also include the output, processing, analysis, and storage of data16.
Consistent definitions and protocols21. In order to minimize the number of ‘translation steps’ in the LAS – where errors can be made in every step – consistency in definitions is essential for automation. Consistent protocols make the work as repetitive as possible, and let the automated systems thrive.
Traceability, integrity checking, and error management11. Even in the most perfectly designed system errors can still occur. Without lab personnel monitoring the process, the automated system acts as a ‘black box’, and can easily obscure these errors. Still, this does not have to be a problem, as long as there is an extensive logbook (in the LIMS) of sample handling steps, including the documentation of errors. As a result, either the system can be improved, or the outliers in data can be easily explained. Operational data indicating component performance is an important factor to be checked, since a sub-optimally functioning device can still indicate execution of all required tasks, but the performance could be inadequate22.
Redundancy in system components and consumables1,20. As indicated in the ‘Cons of total laboratory automation’, a component breaking down without sufficient backup can be detrimental for a process. The same holds for a component running out of consumables. By far the most important system component to have a backup for – but maybe also the easiest one to forget – is the power supply for the entire system.
Fast and reliable customer support20. Even if a component breaks down that is no problem, as long as the downtime can be minimized. Fast and reliable support from the manufacturer, and maybe even an on-site maintenance technician, can get the system up and running as quickly as possible.
Flexibility in combining automated modules into a total automation solution4,5,23. When a fixed total automation solution is purchased, any changes to the experimental setup might render the fixed system useless. Combining multiple modules provides flexibility and, consequently, a long-term gain. However, it can take time to optimize and validate the connection between all separate units21. This short-term investment of time and money needs to be overcome and may need to be (to some extent) repeated each time there is a change to the experimental process. When testing a module, it is essential to not only use materials supplied by the manufacturer of the module24: those will generally show the module at its best, but the module should also function properly with materials from other suppliers. The CytoSMART® devices (with the SiLA server or Python-based software as Open API) have the flexibility to contribute to partial or total laboratory automation, as a part of a larger automation system.
Solid metrics to evaluate automation effectiveness4. When only considering human experience, it will be hard to determine whether all investments in an automated laboratory paid off. Solid metrics, including the combinations of time it takes to process a sample, costs per processed sample, or a number of samples where no repetition of measurements is required can clearly indicate whether the automation is successful4. If not, the metrics will also provide insight into the steps with room for improvement.
Sectors for successful laboratory automation
Certain sectors within life sciences inherently meet more requirements for successful laboratory automation. Generally, large research groups and organizations with high-throughput research that adhere to strict and optimized protocols will benefit from lab automation solutions the most. These can be found in (but not restricted to)16:
- Hospitals and clinical laboratories, but also veterinary and forensic laboratories
- Laboratories of cell biology manufacturers, as well as pharmaceutical and biotech companies
- Large academic laboratories
- High-risk laboratories (biological safety level 3 or 4)
If a single optimal automation solution for all laboratories would exist, it would have been applied in all laboratories around the world a long time ago. However, there is no one best solution, but there is an optimal solution for every single case. Even if total laboratory automation is not suitable for an application, partial laboratory automation and otherwise machines and devices can be. The most important question is always what processes a laboratory performs and needs, and an open discussion with the employees should be held about the optimal solution: because, in the end, there is still a human that presses the button to start an automated laboratory.