Webinar Highlights: Analytical Methods for Semi-Targeted Screening of Pharmaceuticals and Illicit Drugs in Complex Environmental Samples

Dr Leon Barron, Lecturer in Forensic Science, King's College London

Due to target analytical approaches, which detect just a small number of compounds, the true breadth of pharmaceutical and illicit drug residue contamination within an aqueous environment can often be underestimated. High resolution mass spectrometry, in combination with liquid chromatography, can be used for semi-targeted screening of waste- and river-water.

In this interesting webinar, Dr Leon Barron, Lecturer in Forensic Science at King's College London, uses a number of case studies to demonstrate the semi-targeted screening of contaminants in waste-and river-water, the prediction of chromatographic retention time, and how large datasets can be managed with post-acquisition data-mining. Read on for the summary of the Q&A session of the webinar, and if you missed it, watch the free, on-demand webinar here.

The application of the semi-targeted approach is mostly for parent-active pharmaceutical ingredients. Can this be used for metabolites too?

Yes. What we've done recently is expand the number of compounds that we have included in our ANN training data set to well over 500, in collaboration with a group in Spain. What we found is that we can apply ANN to the elucidation of some metabolites, by using the chemical structure you think it might be, to generate descriptors and subsequently a predicted retention time, which will hopefully work in a broad scope screening method for metabolites as well.

As previously mentioned, we are also doing this for biological samples. In particular, we've focused on Gammarus pulex, a river invertebrate, and we hope to elucidate at some stage the metabolic profile of these organisms, as well as the metabolism of these drugs.

Why do you not use neutral pH for the stability, and why do you change the pH to pH 2?

That was an executive decision from the start. It has been shown for some illicit drugs (which were a large component of our compound set) that they are more stable if you acidify then freeze them.

Will the ANN work for other types of compounds, not just pharmaceutically-related compounds?

I think pharmaceuticals are interesting because they are ionizable compounds, compared to other compounds, where other, more linear, retention modeling work has been done. We've recently had a look at some modeling tools to apply ANN to pesticides, herbicides and energetic materials. For the latter, not all of these are ionized compounds and they also contain a lot of different chemical structures, but so far it seems to be working, albeit on a smaller data set of compounds. In theory, therefore, ANN should work on a range of compounds of different chemical make-up.

Can neural networks be applied to other things associated to the method? For example, could you predict the SPE recovery?

Essentially you could, as ANNs can predict a number as an output. For example, we've already predicted things such as sorption potential to sludge using neural networks. Specifically relating to the method, you could look at predicting the peak widths or the SPE recovery. That is a good example, as there is a weakness with the SPE recovery in recovering all compounds, and it would be nice to predict which ones we definitely wouldn't see using this method. We accrue a number of recovery percentages that we can attribute to a certain compound type, so it would be exciting to try out whether they can be predicted.

What do you see as the best potential combination of SPE sorbance for application to boost broad recovery?

We deliberately started off with a single cartridge approach. In my experience, most methods focus on single cartridge approaches. They generally use a mixed mode cartridge, which may or may not include an ionizable functionality (e.g. so a strong cation exchange mixed with some polar or non-polar functionality, or the opposite). Once you've acidified your sample – which protonates the acids (i.e. making them neutral), keeps the neutral species unionized and ionizes the bases – you may want to use both a cation exchange cartridge along a broader scope mixed-mode cartridge incorporating a polar and non-polar functionality.

What we don't know yet is the best combination of cartridges in terms of arrangement. That is, can you can do this in a parallel-type extraction approach with two separate cartridges, which may then be combined at the end for analysis, or whether we should do it in sequence. I would probably start there, with the generic sorbent we used here, followed by a cation exchange sorbent, but we have to work through the data and see what difference that makes. We're working through this at the moment to see if a dual cartridge approach actually gives us any extra recovery, without exacerbating matrix effects to the point of reducing sensitivity of the method beyond a practical level.

Why is there such a retention shift observed with this LC method when running real samples in comparison to ultra-pure water?

Initially we were quite worried about the effect of this on our method performance. The main reason is the matrix, but there also seems to be some disparity in the SPE concentrations factors used in the literature. A few studies use concentration factors of ~200 but some go up as high as 2000, which really concentrates up quite a bit of matrix. Here, we decided on somewhere in the middle, settling for a 1000 fold concentration factor for our waste water and river water, but it will still concentrate up quite a bit of matrix, and as we observed, this affected the retention behavior of our analytes. I can only assume that is caused by the matrix adhering to the stationary phase, or perhaps there is some interaction in the mobile phase with components of the matrix, preventing retention. Generally we saw an increase in retention as opposed to a decrease in retention, which was strange, but I think that it is mainly due to the SPE concentration factor used.

There is an argument against our approach too − if you increase your SPE concentration factor, you can also increase your ion suppression effect, so it would make sense to go for a smaller concentration factor. The simple reason for our approach was that this was intended to be a semi-targeted or ‘suspects screening’ method, not a targeted method, so actually some co-extraction of the matrix components is what we are trying to do. You've got to find a fine balance and there is lots of work still to be done to really agree on what is an acceptable concentration factor for the SPE method, and hopefully we'll solve that one too!

If you missed this webinar, you can still watch it here.