Lab Workflow

High Throughput Formulations: Some pitfalls and suggestions when  choosing or designing a system

Interest in applying High Throughput technologies to formulations has been growing in recent years. Use of automated 'high throughput' systems, techniques and tools offer formulation chemists opportunities for improving experimental throughput, precision and accuracy. In turn, these improvements allow the chemist to better explore the experimental space. Of course, it is possible to employ manual techniques and labour in a 'brute force' approach but the effect on human morale and patience will eventually affect the quality of the experiments.

Several suppliers are now offering tools and systems specifically designed for the formulation chemist; of note are the systems described at a FSTG organized HTFS conference from hte ,ChemSpeed and Zinsser and those from Symyx. In addition, companies like hte, TTP, LabMan, Bosch, ATS, etc., also offer custom systems. Users should be cautious about the applicability of such systems, or those from any other manufacturer, since, in addition to the opportunities of improved experimental performance offered by high throughput tools, there are significant opportunities for selecting the wrong tool and ending up with an expensive 'white elephant'

This article is provided to give users a starting point when selecting or designing their own systems.

Typical problems include slow tools or tools which are presented with high numbers of samples which create bottlenecks in the workflow, tools that exhibit imprecision in the performance of the task or even tools that are not capable of the task it was specified to perform. In addition, the servicing tasks of the systems may not have been considered and you may end up spending too much of your time performing set-up and after-run procedures such as cleaning.

Antomaxion's approach has always been to start with the goal in mind. If you are a technologist close to the techniques, it is very easy to get excited about the technology and to

 lose sight of the experimental goal. It is important to realise that the entire sample that is prepared should be destined for measurement and only sufficient for that purpose should be prepared. If possible, scaling down the sample size will allow more samples to be prepared from the same amount of starting material, allow more samples to fit on a system and produce less waste. The size of sample prepared will therefore depend on the amount of sample needed for all the tests and also on the capability of the tools used in the preparation including precision and accuracy of dispensing, dimensions of the processors and losses between processes. It should also be noted that if more sample is needed on the odd occasion and since the system can by definition prepare samples reproducibly, more sample can be prepared on demand.

Having established how much sample must be prepared, it is desirable to maximise the availability of each analytical device by minimising the number of samples carried forward in the process, minimising the analysis time and minimising the use of methods which involve a cleaning cycle.

An example of application of these simple rules would be to measure pH and make adjustments during a homogenisation step rather than adjusting pH in a post-homogenisation step. Another might be the monitoring of a process using IR spectroscopy which is a fast measurement in itself, but if performed through the vessel wall would also not require any cleaning. If the characterisation step of rheometry were included early in a workflow, this slow and detailed measurement would cause a bottleneck and decrease the efficiency of the whole system. By compromising on the measurement by replacing this characterization step with the screening of viscosity, good throughput can be maintained. In addition, certain viscometers can yield some rheometric information such as shear thinning or thickening behaviour and eliminate the need for rheometry even at a later stage.

Unlike during the measurement steps where there are opportunities for eliminating samples, when choosing tools to perform the actual processing, such as mixers, grinders etc., the goals must be to provide multi-functionality to allow more processes to be performed in the same space, minimise cleaning to ensure maximum utility of the tool, and maximize the efficiency of the tool so that the tool is occupied for the minimum amount of time. Some examples of tools that provide these efficiencies are mixers which both mix with paddle blades and homogenizer rotors but also allow the use of measurement probes, and milling devices that either comminute within a closed vial and therefore don't require cleaning, or provide vigorous mixing by the same action.

Choice of appropriate dispensing technology will depend equally on the number of dispenses during a run and how often a material is changed as well as the performance capability of the dispenser. This is true, even for so-called normal liquids where the choice is between a probe connected to a bottle and pump or disposable pipette tips. The former is most suited to dispensing a single fluid from the bottle but can also be used to transfer fluids between vessels on the robot deck, which is the normal task of the disposable tips, using the fluid from the bottle as the hydraulic fluid. When very viscous materials are to be dispensed (viscosities in the 10 000's of cP) the selection of a tool becomes even more complex and an understanding of the frequency of the dispense and how many runs a device will be used for is critical. If, for example a glue like material is to be dispensed only a few times from a bulk container, there are many devices available for performing the dispense, but in this case, the manual set-up and cleaning after use could be more time consuming than performing the few dispenses manually! In this case, the selection of a tool such as those from TriDak, where the wetted path can be removed and replaced quickly and easily, is essential. In the cases where a few dispenses of many different materials is needed, they can be transferred from a vessel on the robot deck using a disposable tip. I have even had a robot dispensing multiple samples of melted pich in this manner, using heated disposable glass pipette tips and a Zymark robot (yes, this was a while ago).

When dispensing solids, the number of standard technologies available are limited to vacuum canula for transfer between containers or dispensing from hoppers or vials with appropriate technology and the selection will depend on the required precision and accuracy, the nature of the source container and the size/frequency of the dispense. Such devices can be obtained from companies such as ChemSpeed, Zinsser, Autodose, Mettler Toledo AutoChem and others.

Once all the individual processes and tools have been defined, the way in which these are integrated together becomes critical. They could be devices that stand alone and are linked only through the user, or they may be fully integrated. Most users are leery of committing to building a full scale automated high throughput formulation system without an estimate of the eventual throughput so that the cost/benefit can be calculated. In addition, users often want to build a basic system which can be expanded to include other capability or increase throughput. In the same way that the 'LabSmart' audit of a lab by the LGC can yield insight into where automation can best be applied, a study of the collection of the selected tools into virtual systems, figure 2, is highly recommended. This allows the throughput for various recipe's to be estimated, figure 3, bottlenecks to be found and eliminated and the effect of new recipe's or additional tools to be determined, figure 4.

All these considerations imply that successfully building a High Throughput system for formulations needs careful thought. I would point out that most problems will occur when the unexpected result occurs: a formulation is made with unusual properties or an operator makes an error. There is considerable experience in the Formlulations community and it is growing all the time. If you are considering High Throughput it would certainly be worth attending the next HTFS conference or talking to other about their experiences.

Although this article is written assuming that High Throughput means the use of automation and robotics, all these rules can be applied to  non-automated workflows.