Hello everyone, my name is Ben and I’ve recently joined JMP as an intern focusing on developing content for the Functional Data Explorer platform that is available in JMP Pro 17 (you can read more about me here). I’m currently studying my PhD in Chemical Engineering at The University of Manchester with a focus on sophorolipid biosurfactant production. This has given me a lot of insight into the potential data/needs for processing data in the early stages of development of bio-products. The case study here describes my work in developing an analytical method to correctly detect sophorolipid biosurfactants, using functional data analysis and functional DoE to guide my experimental protocol.
What do you like about this case study?
- It uses functional data analysis and FDoE – can show both major components of the Functional Data Explorer tool
- The data-set is real and based on actual development work
- You can show very compelling visuals - "wow" factor.
- You can utilise the FDE platform to easily do pre-processing (alignment, filtering)
Who is it for?
- Anyone involved in Analytical development – including chemical, biological and pharmaceutical industries.
- Anyone from R&D should be able to see how it could translate to their situation.
How should I use it?
- You can use it to demo the Functional DoE tool in JMP Pro. The attached data table has saved scripts for each stage of the process or you can follow the case study details below to recreate the work.
- You can use it to show the ability to simplify data presentation in the Functional Data Explorer platform.
What platforms can we use?
- Graph Builder
- Functional Data Explorer
Background
Sophorolipid biosurfactants are produced by the fermentation of a yeast with sugar, oil and nitrogen that are composed of a hydrophilic head and hydrophobic tail. As with a lot of bio-products there is massive variation in these structures in a single fermentation in terms of chemical bonds, chain length and functional groups (to name a few). For analytical scientists this provides a challenge in terms of separating these compound efficiently when using analytical tools like high performance liquid chromatography (HPLC).
In this case study, I was faced with the challenge of trying to produce a method capable of quantifying sophorolipid biosurfactants with HPLC-UV, which had not been previously done. Unfortunately, my initial method was not able to separate two sophorolipid congeners (C18:1 and C17:1) from another with HPLC. They only differed by one carbon, meaning they had similar absorbance (the wavelength of light they absorb) and retention time (the time they appear on a HPLC spectra). I couldn’t do anything to change the absorbance, but I could adapt my HPLC method to change the retention time of the two products so they separate better. Peak separation is vital in analytics as the peak area is used to calculate the concentration of a given compound – without ‘standalone’ peaks this calculation is inherently flawed and can lead to inaccurate readings. The example described here represents a common challenge in analytics and has relevance to a wide variety of industrial and research groups.
Aims
- To optimise the HPLC settings in order to effectively separate the C18:1 and C17:1 peaks
- Ensure that the peaks are sensitive (no major loss in peak height) and specific (no broad peaks)
Method development – Initial DoE application
To optimise the HPLC method, I applied a simple DoE that adapted the temperature (30, 35 and 40oC) and flow rate (0.7 and 1.4mL/min) of the method. The same concentration of sophorolipid sample was tested in every iteration and all other conditions were kept constant throughout.
The output for the HPLC is a chromatogram – this measures the intensity of absorbance at a wavelength (198nm) over time, each peak represents a compound coming off the HPLC and hitting the detector – the longer each compound interacts with the HPLC column the longer it takes to reach the detector, causing a later retention time. Most samples that are put on HPLC’s contain hundreds of compounds (including your compound of interest) producing a very busy graph that makes it impossible to discern the specific differences between each setting by eye.
To further complicate matters, the change in settings (particularly flow rate) meant that our peak of interest (for the C18:1 and C17:1 forms) did not have the same retention time. This not only makes it hard to pick your peak out, it also makes it very hard to produce a comparison graph when you have numerous runs without the graph becoming over-complicated.
Application of the Functional Data Explorer and FDoE tool
To explore the changes to the chromatogram the Functional Data Explorer tool was applied (Analyze > Specialized Modelling > Functional Data Explorer). The settings were applied as above.
The initial data set showed all of the chromatogram which is very busy, whereas the region of interest was in the C18:1 and C17:1 peaks which occured between 22-30 minutes. The time range was reduced from 22-30 minutes using the Commands>Filter X tool.
In order to gain anything meaningful in the comparison the separated peaks (changed due to the change in flow rate) needed to be aligned. As the peaks of interest had the biggest intensity in the range of interest (as is typically the case in HPLC for specific compounds) this was done easily with the Align>Align Maximum tool. Another Filter X was used to reduce the range to just the peak of interest (-1 to +1 aligned time). Peak alignment is used a lot in analytics – particularly in LC-MS and HPLC process development and the ease of applying it with the FDE tool is very useful for analytical scientists.
Now that the data was ready for analysis, a Wavelet model (Dropdown > Models > Wavelets) was applied as this best fitted the chromatographic data. To be able to profile the effect of changing the temperature and flow rate the Wavelet DoE (Dropdown > Wavelets DoE Analysis) was used to profile the chromatogram.
Looking at the change in flow rate (Left), the FDoE profiler easily demonstrates how the change in flow rate causes a sharpening of the peak (shown by the increase in the peak intensity value) and the appearance of a second peak (the C17:1 compound). Further to this, the changes in temperature (Right) further improved this separation and sharpening of the peak. From this profiler, it became clear that the peaks were separating but not sufficiently – this indicated that more experimental runs were required.
Further runs
From the output of the Wavelet functional DoE profiler 4 additional runs were performed – 2 looking at increased temperature (45 and 50oC) and 2 with increased flow rate and temperature (45oC, 1.6 and 1.8mL/min). The chromatographic data was added to the previous dataset and the Functional Data Explorer tool used to produce a Wavelet DoE profiler as described above.
The FDoE/Wavelet profiler was able to demonstrate that flow rate had dimishing returns for the HPLC – it did not improve peak separation greatly at higher flow rates and led to a decrease in the peak height (meaning reduced sensitivity). As there was no significant improvement in separation the flow rate was kept at 1.4mL/min in order to sustain the column (higher flow rates can be detrimental to HPLC’s so this was able to balance risk vs. reward).
When looking at temperature, the profiler was able to show significant improvements in both the peak height (improving sensitivity) and the separation of the second peak
Why does this matter?
From a customers perspective, the need for FDE and FDoE can be explained by a simple question:
Would you rather try to explain the differences with this?
Or this?
The use of the FDE platform can provide significant time savings, streamline result production and allow for easier discussions with non-SME’s when discussing your results. When this is placed in tangent with FDoE you get a powerful tool that can identify the whole change in functional data/profiles, rather than using single points (i.e. peak area in a HPLC, peak retention time). The case study here demonstrates the use of FDE to inform the development of the HPLC method to produce an optimised method.
