Lhasa Limited shared knowledge shared progress

Application of quantum mechanics and/or structural fingerprinting for the prediction of glutathione conjugation site-specificity

pdf file

Glutathione conjugation of drugs or xenobiotics is an important metabolic process with repercussions (both positive and negative) for chemical toxicity. However, the potential for such reactions may go unrecognised and the specific site of reaction may influence the toxicity in vivo. Quantum mechanical calculations have long been seen as a means of identifying electrophilicity and more recently as a means of distinguishing reactions with hard and soft nucleophiles. However, enzyme mediated reactions complicate predictive modelling of metabolite products. In this work we look at the contribution that quantum mechanics can make over and above the use of sub-structural fingerprinting.

An in-house dataset of glutathione conjugation reactions, generated from an earlier careful extraction from published literature, contained 1038 reactions for 608 starting molecules. Using NWChem 6.6, a range of molecular and atom-based descriptors were generated including HOMO and LUMO energies and their atomic populations, Wondrousch and conventional local electrophilicities, and Fukui reactivity indices. Two fingerprints were also calculated; firstly MACCS keys, which relate to functional groups and may thus be useful in a reactivity analysis, and an atom typer fingerprint developed in-house, modified from the Ghose and Crippen atom types.

Using the experimental data and calculated descriptors, statistical models for the prediction of the sites of conjugation were developed using a Support Vector Machine (SVM) methodology. The performance of models globally and for individual reactivity classes have been considered. Comparison was also made with predictions using Meteor Nexus (Lhasa Limited). The local fingerprint was originally included as a counter-hypothesis to the use of quantum-mechanical descriptors, but showed better performance, alone, than the latter. However, the best models for many classes included the quantum-mechanical descriptors indicating that these are useful when searching for models regardless of computational cost. The strengths and weaknesses of significant descriptors are considered with a view to improving performance and giving mechanistic interpretations of models.


Presented by Dr Martin Payne at the Joint UK-QSAR and Molecular Graphics and Modelling Society conference in April 2018.



© 2019 Lhasa Limited | Registered office: Granary Wharf House, 2 Canal Wharf, Leeds, LS11 5PS, UK Tel: +44 (0)113 394 6020
VAT number 396 8737 77 | Lhasa Limited is registered as a charity (290866)| Company Registration Number 01765239 (England and Wales).

Thanks to QuestionPro's generosity, we now have survey software that powers our data intelligence.