16:00 - 18:00
Room: Poster Area - Poster Shed
Poster Presentation
Building a prediction model to distinguish Saint John’s wort samples based on their 1H-NMR profile
Francesca Scotti 1, Birk Schuetz 2, Andrea Steck 2, Michael Heinrich 1
1 Research group Phytotherapy and Pharmacognosy, UCL School of Pharmacy,, London
2 Applied, Industrial & Clinical MR Market Division, Bruker BioSpin GmbH,, Rheinstetten

Saint John’s wort (SJW, Hypericum perforatum L.) is a widely spread and well-known medicinal plant. Previous studies conducted in our group (Booker et al. 2018; Scotti et al. in preparation), using a combination of NMR metabolomics and HPTLC, have shown some notable differences exist between materia prima samples’ of different origin, some of which are ascribable to geographical provenance,. These “variations” are not represented in SJW European Pharmacopoeia (EP) standard, nor mentioned in its monograph’s HPTLC description.

80 samples of SJW crude drug material belonging to our previously acquired collection (Scotti et al. in preparation), were analysed by 1H-NMR at Bruker (Rheinstetten, Germany). The NMR results were used to build a statistical model using PCA/LDA combined with Monte-Carlo-crossing for assessing the possibility of predicting samples belonging to distinguishable chemical entities, based uniquely on their 1H-NMR spectra. The trials initially looked at possible distinction based on geographical origin with special consideration for samples from China, known to have a unique fingerprint (Booker et al. 2018, Scotti et al. in preparation). Further trials aimed at finding a predictive model for pharmacopoeial compliance. The resulting models had limited ability to predict geographical provenance but had stronger ability indicating which samples are of Chinese origin (92% correct estimate). A significantly increased capability to determine compliance to the EP standard (up to 99%) was obtained.

One limitation of this study was the circumscribed number of samples belonging to each geographically distinct area, which derived in the necessity to cluster adjacent areas as single groups with larger number of samples, which might not necessarily share the same chemical profile. Therefore, we hope to implement the sample collection to provide stronger bases and more significant geographical boundaries for a more comprehensive predictive model.

References:

Booker et al. Phytomedicine (2018); 40:158-164

Scotti et al. MS in preparation


Reference:
Poster session-PO-111:
Session:
Poster Presentation-2
Presenter/s:
Francesca Scotti
Presentation type:
Poster presentation
Room:
Poster Area - Poster Shed
Date:
Tuesday, 28th August, 2018
Time:
16:00 - 18:00
Session times:
16:00 - 18:00