Landmark Literature: 2015
We ask an impossible question as 2015 comes to a close: which piece of literature stood out from the crowd and showed the greatest potential for pushing the field of analytical science forward?
Davy Guillarme, Hans-Gerd Janssen |
By Davy Guillarme, Senior Lecturer, School of Pharmaceutical Sciences, University of Geneva/University of Lausanne, Geneva, Switzerland
Let me begin by setting the scene for the paper I’ve chosen. In the last few years, people are regaining interest in supercritical fluid chromatography (SFC) at the analytical scale. This is largely due to the commercialization of new generation SFC systems. These systems offer enhanced sensitivity, robustness, and quantitative performance. In addition, they present a better compatibility with modern stationary phases, such as those packed with sub-2 µm fully porous particles or sub-3 µm superficially porous particles.
In our laboratory, we have evaluated ultra-high performance SFC systems (UHPSFC) and have found that both reliability and performance has significantly improved. However, the extra-column band broadening observed on such instruments is non-negligible because of the long, wide connection tubes employed and the UV cell, which represents a volume of up to 8 µL. Therefore, it would be very interesting to use short 2.1 mm ID columns – today’s standard in ultra-high performance liquid chromatography (UHPLC). Not only would such columns reduce band broadening, but solvent and sample consumption would be reduced, making UHPSFC a much greener strategy than UHPLC. In addition, the mobile phase moves at relatively high linear velocity under UHPSFC conditions, so narrow bore columns would be appropriate to stay within the flow rate range of current instrumentation (upper flow rate limit of 4 to 5 mL/min on UHPSFC systems). Unfortunately, short 2.1 mm ID columns are not well suited to UHPSFC because of the significant loss of efficiency.
Now to my landmark paper for 2015: “Understanding and diminishing the extra-column band broadening effects in supercritical fluid chromatography” – by Ruben De Pauw, Konstantin Shoykhet, Gert Desmet and Ken Broeckhoven. The authors identified and quantified the different contributions to extra-column band broadening, including the influence of sample solvent, injection volume, detector cell volume, and tubing volume. Rather than calculating the extra-column variances for different system geometries, which is not an easy task when using a compressible fluid, the authors decided to evaluate the change in performance by tracking the plate count for a few model compounds with different retention factors. After complete optimization of the UHPSFC instrumentation, the plate count was drastically increased and the optimized system was found to be fully compatible with narrow-bore 2.1 mm ID columns.
Some of the findings in this study were particularly relevant and unexpected. First, the authors have shown that, when using a commercial modern SFC system, minimizing the tubing ID from 250 µm to 170 µm and then to 65 µm was not a good strategy for reducing extra-column variance, since the achieved plate count was not modified on a standard system. These findings contradict what is generally observed in LC, where this parameter is one of the most important for improving peak shapes. The different behavior can be attributed to the higher molecular diffusion coefficients in SFC, leading to a lower contribution of tubing to extra-column band broadening, thus allowing the use of larger ID tubes compared with LC. Secondly, the UV cell volume was reported as an important parameter for minimizing peak broadening, and so using UHPSFC systems with a UV cell volume of <1 µL is highly relevant. However, you need to take care when reducing the UV cell, since turbulent flow conditions are much more common in UHPSFC than UHPLC and can cause an unexpected increase in pressure in the narrow flow path of the low dispersion flow cell. Finally – yet importantly – the authors prove that the most important parameter to achieve suitable peak shapes on narrow bore UHPSFC columns was the nature of the injection solvent and its volume. The sample diluent should indeed match the mobile phase in terms of viscosity, elution strength, and polarity, while the injection volume should be minimized.
In conclusion, this study provides some useful tips to chromatographic instrumentation vendors that should help with improving their UHPSFC systems. I expect the next generation of commercial UHPSFC systems to work optimally with narrow bore 2.1 mm ID columns. For this to happen, UV cell of less 1 µL in volume that withstand pressure up to 400 bar need developing, and we need autosamplers that offer precise injection at very low quantities (< 1 µl). However, reducing tubing ID below 120 µm doesn’t appear to be a useful strategy, since it increases the system pressure, with limited benefits in terms of achievable plate count.
Davy’s Landmark Paper
R De Pauw et al., “Understanding and diminishing the extra-column band broadening effects in supercritical fluid chromatography”, J Chromatogr A, 1403, 132–137 (2015). PMID: 26054561.
The Birth of Data Fusion
By Hans-Gerd Janssen, Science Leader, Unilever Research and Development, The Netherlands
Science is about generating new knowledge and understanding. To do so, we develop hypotheses and then perform experiments to either accept or reject them. The process of hypothesis testing centers around data; indeed, the data are collected with a specific purpose in mind: to test the hypothesis. However, data collected for one purpose can be valuable for numerous other situations, especially given the dramatically improved data-density of our current analytical measurements. And with the improved systems we have for storing and transferring large data sets, we are now ready for recycling, re-using, and reinterpreting data. To allow this, the data must be aligned, which is defined as “made laboratory independent”. My choice of landmark analytical paper describes a method to do so (1). More importantly, it expresses and emphasizes this idea of not just using (selected) data for a single purpose, but to store and combine them, and use previously measured data to speed up hypothesis testing, possibly with very different purposes in mind.
The paper was written by a team of analytical experts from the Netherlands Metabolomics Centre (Leiden) led by Thomas Hankemeier alongside co-workers from several Dutch academic hospitals wrote the paper. And although the work focuses on metabolomics, the recommendations made in the article reach far beyond into other fields. It is a plea for making data amenable to data fusion – merging of data to a seamless data set – and describes the precautions that need to be taken to allow doing so.
Fusing datasets is easy when the data are absolute or quantitative. The analysis of a given sample should give the same absolute concentrations irrespective of where it is measured. But modern instruments deliver much more data than just those for a few target compounds that are normally quantified. The paper examines those additional untargeted data, the much high number of compounds for which only relative, semi-quantitative information is available.
The authors describe the use of re-measuring a subset of samples from different studies to obtain “transfer models”. Application of a transfer model then enables matching and integration of semi-quantitative profiling data between different studies and sample sets measured at different times and/or in different laboratories. Alternatives for a possible lack of transfer samples (for instance, due to limited sample volume) are given. Reference samples such as those available from the US National Institute of Standards and Technology (Gaithersburg, Maryland, USA) are a first option, but other solutions are also included.
Ideally, after completing each set of measurements a set of transfer samples is analyzed. This set provides the bridge to a new study, and enables integration of all the studies performed. Reanalysis of approximately 6–7 percent of the total number of samples was found to be optimal for establishing the transfer model, but it is possible to do less.
There is still a long way to go before we can simply look on the Internet for the data we need to test our hypotheses. I sincerely believe, however, that in the future we will not only Google for information, but also for data. Data transfer strategies such as those developed in the paper will help that happen – just like good laboratory practice in putting data on the Internet.
Hans-Gerd’s Landmark Paper
AD Dane et al., “Integrating metabolomics profiling measurements across multiple biobanks”, Anal Chem, 86, 4110–4114 (2014). PMID: 24650176