Cost Versus Payback

The goals for improvements in sample prep include:

• Overall cost reduction, including labor and materials • More consistent preparation
• Higher productivity—more samples processed, which may be coupled with automated instruments
• The ability to work with hazardous materials • More extensive analysis—work that might be too expensive for manual efforts such as statistical experimental design and high-throughput screening
• The potential for 24/7 operations

For any ROI equation, there are two sides to consider. The first is what you want out of it, which includes some or all of the points above plus metrics—what level of performance are you looking for, what are you willing to spend, and what is the schedule requirement for implementation? You also have to evaluate the alternatives to automated systems, which include increasing head count or outsourcing work for comparison. Those last points would have to include an understanding of whether the need is a temporary spike in testing throughput or a long-term requirement; it is going to take time to implement a solution, and you don’t want it coming online as the need evaporates.

The other side of the equation covers the costs. They include the development of the user and system requirements, a feasibility study, and prototype work, followed by the actual design, implementation, documentation, testing, validation, and user education. Given a set of requirements, the next major step is the feasibility study—this is going to provide the basis for the go/no-go decision on the project and guide the design effort. The first step in that study is an evaluation of the sample preparation procedure, the underlying process of the system.

Unless the process is specifically designed for automated implementation, the process is going to have to be analyzed to see what it will take to make it suitable for semi-automated or fully automated work. If there is a make-or-break step in the project, this is it. Lab procedures— and given the readership of this magazine, we’re covering a broad spectrum of procedures—describe the science behind the work and what steps people should take to carry it out, relying on the individuals’ intelligence to fill in gaps or make things work. The first item that has to be determined is whether you are using the current, up-to-date description of the process, including any undocumented workarounds, temporary fixes, etc.

Next, the feasibility analysis has to evaluate whether there is anything about the process that precludes automation. This would include working with objects or materials that depend on human dexterity and might not be usable with robotic systems. If that is the case, is it possible to modify the equipment or process to make the automation work without altering the science itself ? Another consideration is whether the process can be optimized—this may be necessary in order to meet performance goals.

Early robotic sample preparation system implementations mimicked human activities, carrying out the same steps, one at a time, that people did. This removed people from the system (one goal) but often didn’t process more samples per hour, although it did provide a means for three-shift operation. Process optimization may require a change in how the process takes place, as long as it doesn’t compromise the integrity of the results. Examples of this can be found in life science’s implementations of assays and screening using microplates. Microplates can carry from as few as six to as many as 3,456 wells, although 96- and 384-well plates are more typical. Each well—essentially a miniature test tube—carries one sample, blank or standard for analysis, and allows for parallel processing of the plate’s well contents. That allows for very high throughputs for colorimetric work or other analyses. One question is whether this technology can be applied to a wider range of lab applications. The plates can be made from polystyrene, polypropylene, cyclic olefin copolymers, and glass. Given the availability of microplate-handling equipment that includes washers, liquid additions, sealers, and readers, it may be possible to apply microplates to analytical work outside life science applications. There is one caution: the small sample size of the plates’ wells (each well in a 96-well plate is 360 μl—the volume may vary for special configurations) requires highly homogeneous samples; that may not be a problem with liquid samples but could be with solid material that has to be dissolved. Solvent leaching of additives (antioxidants, etc.) is a potential problem when working with small volumes with large contact areas; you may not dissolve the material, but small molecules may migrate from the holder into your sample.

The high throughput of life science assays via microplates is a result of standardization of microplate geometry. Sample preparation methods that are based on standardized sample containers will reduce the effort in automating systems. This is particularly true of autosamplers that use standardized vials. The Agilent 7693A ALS, for example, is an autosampler that can be programmed to process samples prior to injection into a gas chromatograph. It has the ability to carry out liquid/liquid extractions, small-volume sampling, reagent and standard additions, heating, mixing, reconstitution, dilution, aliquoting, and bar code reading programmed through a chromatographic data system. The use of standardized components reduces development costs, increases the likelihood of success, and permits reconfiguration of systems as needs change.

If your application requires a custom-designed solution based on individual components (robotic arm, sample holders, etc.), the cost of development can increase significantly. The service life of the sample preparation system has an impact on your choice of standards-based or custom solutions. This is one area where the formation of an active user community can benefit lab work. Unless you believe that having a particular automated sample preparation system provides a competitive advantage, consider having a partnership with companies with similar needs to jointly develop automated solutions. This should reduce the cost of development and provide a more robust system with better support. It may also provide enough market justification for an equipment vendor to step in and develop hardware to fit the application. Standards, in some cases, do not have to be universal but can be industry/community/ application-specific.

A sample preparation system’s service life can be measured two ways: by the calendar and by the number of samples. Sample volume (samples per day for the expected service life of the system) is the key factor. The ROI comparison between manual (including outsourced) sample preparation methods and automated systems is going to depend on the volume of samples needed to justify the cost of development. The calendar component comes into play when the system is going to be in use long enough to require software or hardware changes or upgrades—software update schedules will usually occur on an annual basis (but vary from one vendor to another), and hardware component changes may occur on an 18-to-24-month schedule (using computer hardware as a guide), about the time people start looking for more performance from a system. The potential for system changes is going to have to be factored into the system design; the upgrade process can cause systems to fail, particularly if developers have modified underlying system software that is compromised by the upgrade. Good development practices should take this into account, but they often require more effort. Systems based on standardized components are more likely to avoid this problem, requiring shorter development schedules and reduced development costs.

One point noted earlier was the use of a prototype system in the feasibility phase of the work. This stage is often ignored; once something that “works” is put together, it is left as is. There are important questions that have to be asked, and two of them are “If we had to do this again, how would we change/improve it?” and “What did we learn from this?” The answers to those questions separate a well-designed system from something that will have to be fixed later on. There is an unfortunate phrase that comes up in a project when schedules and budget are getting tight: There is never time to do it right, but there is always time to do it over. Doing it right is the only way that makes sense. The alternative is a compromised process that may not yield the quality of results that is expected, and you may not find out about it until a significant problem develops.

Automated sample preparation systems can address special needs that can skew the ROI evaluation. If you are working with hazardous materials or samples/reagents that require special handling or controlled environments, automated systems may be better from a safety standpoint. Centralized sample processing using pneumatic or rail-based sample delivery to instruments can make this feasible. The change to the ROI analysis is the cost of the system versus the safety and centralized material-handling benefits.

What has been described so far is a scientific manufacturing approach to laboratory sample preparation. We are looking at scientific processes that take incoming materials and transform them into samples ready for analysis or, ideally, continue the process to include automated introduction of the samples into instruments. The steps that have been outlined (cost considerations, feasibility analysis, process optimization, consideration for implementation methods) are common to any production process. They aren’t within the experience of most lab or IT personnel. Where do you get help with doing this work? Find people with chemical process engineering, bioprocess engineering, and/or mechanical engineering backgrounds; it is a matter of applying their skills to lab work. In many cases it is just a matter of scale between typical process engineering and laboratory applications. In addition, they will be in a position to advise you about statistical process control and managing automated systems.

This is a different way of looking at laboratory work, a shift from hands-on experiments and procedure execution to working with systems that do that work for you. That will require a change in people’s responsibilities and education. They will need to understand the process being carried out but also understand the equipment, how it functions, troubleshooting, evolutionary operations (making small, controlled, incremental changes to improve a process), and statistical process control to detect and correct process changes. This will add to the cost side of the equation, but the payback will be significant. Automated sample preparation has the potential to process samples at a lower cost and higher throughput than manual methods, with more control over the quality of the result— less variation in sample preparation. However, the system has to be carefully evaluated, tested, validated, and maintained to ensure that quality.

Automated sample preparation is not always the best approach. It takes time to design and implement, and some processes that are too entrenched in equipment designed for human use may not be cost-effectively converted. It may require new equipment designs. Processes need to evaluated and designed for automated use. If the justification is there, automated sample preparation can be a cost-effective tool in lab work.