XenoSite P450 Metabolism

XenoSite Web provides an implementation of the XenoSite Cytochrome P450 Prediction Models [1]. Cytochrome P450s (CYPs) play a fundamental role in the oxidative metabolism of over 90% of FDA approved small-molecule drugs. As such, predicting the way that small molecule drugs are oxidized by CYPs can aid in the rational design of safe, effective drugs that exhibit high bioavailability. You can try XenoSite P450 Metabolism via the online submission tool. For technical details, please see here.

Please note that XenoSite Cytochrome P450 Prediction Models provide predictions of regio-selectivity (which atoms on a molecule are likely to be oxidized by a given CYP enzyme), but they do not explicity model selectivity (which molecules are substrates of a given CYP enzyme).

If you would like performance measures or technical details about XenoSite, please see our Technical Details

If you would like to use results from XenoSite Cytochrome P450 Metabolism Predictor in your publications, please cite the following paper:

Zaretzki, J., Matlock, M., & Swamidass, S. J. (2013). XenoSite: Accurately predicting CYP-mediated sites of metabolism with neural networks. Journal of chemical information and modeling, 53(12), 3373-3383.

XenoSite Epoxidation

Drug toxicity is often caused by electrophilic reactive metabolites that covalently bind to proteins. Epoxides comprise a large class of three-membered cyclic ethers. These molecules are electrophilic and typically highly reactive due to ring tension and polarized carbon-oxygen bonds. Epoxides are metabolites often formed by cytochromes P450 acting on aromatic or double bonds. The specific location on a molecule that undergoes epoxidation is its site of epoxidation (SOE). Identifying a molecule’s SOE can aid in interpreting adverse events related to reactive metabolites and direct modification to prevent epoxidation for safer drugs. This study utilized a database of 702 epoxidation reactions to build a model that accurately predicted sites of epoxidation. The foundation for this model was an algorithm originally designed to model sites of cytochromes P450 metabolism (called XenoSite) and that was recently applied to model the intrinsic reactivity of diverse molecules with glutathione. This modeling algorithm systematically and quantitatively summarizes the knowledge from hundreds of epoxidation reactions with a deep convolution network. The final epoxidation model constructed with this approach identifies SOEs with 94.9% area under the curve accuracy and separated epoxidized and non-epoxidized molecules with 78.6% accuracy. Moreover, within epoxidized molecules, the model separated aromatic or double bond SOEs from all other aromatic or double bonds with accuracies of 92.5% and 95.1%, respectively. Finally, the model separated SOE from sites of sp2 hydroxylation with 83.8% accuracy. Our model is the first of its kind, and can be used for the development of safer drugs by identifying which drug candidates are likely to form epoxide metabolites.

If you would like to use results from XenoSite Epoxidation Predictor in your publications, please cite the following paper:

Hughes, T. B., Miller, G. P., Swamidass, S. J. (2015). Modeling Epoxidation of Drug-like Molecules with a Deep Machine Learning Network. ACS Central Science, 1(4), 168-180.

XenoSite Quinone Formation

Many adverse drug reactions are caused by electrophilically reactive drug metabolites that conjugate to nucleophilic sites within DNA and proteins, causing cancer or toxic immune responses. Quinone species, including quinone-imines, quinone-methides, and imine-methides, are electrophilic Michael acceptors that are often highly reactive, and comprise over 40% of all known reactive metabolites. Quinone metabolites are created by cytochromes P450 and peroxidases. For example, cytochromes P450 oxidize acetaminophen to N-acetyl-p-benzoquinone imine, which is electrophilically reactive and covalently binds to nucleophilic sites within proteins. This reactive quinone metabolite elicits a toxic immune response when acetaminophen exceeds a safe dose. Using a deep learning approach, this study reports the first published method for predicting quinonation: the formation of a quinone species by metabolic oxidation. We model both one- and two-step quinonation, enabling accurate quinonation predictions in non-obvious cases. On the atom level, we predict sites of quinonation with an AUC accuracy of 97.6%, and we identify molecules that form quinones with 88.2% AUC. By modeling quinonation, the most common type of reactive metabolite formation, our method provides a rapid screening tool for a key drug toxicity risk.

If you would like to use results from the XenoSite Quinone Formation Model in your publications, please cite the following paper:

Hughes, T. B. and Swamidass, S. J. (2016). Modeling Quinone Formation from Drug-like Molecules with Deep Learning (under review).

XenoSite Reactivity

Despite significant investment of resources, around 40% of drug candidates are discontinued due to toxicity, often arising from reactions between electrophilic drugs or drug metabolites and nucleophilic biological macromolecules, like DNA and proteins. The resulting adducts for DNA may cause gene dysregulation and modifications of the code, while protein adducts can disrupt normal biological functions and induce harmful immune responses. Knowledge of the likelihood of adductive reactions would provide critical insights on their probable role in biological dysfunction, and thus, we modeled the reactivity of molecules to DNA and protein along with common screening traps for electrophiles, i.e. cyanide and glutathione. A deep convolution neural network was trained on literature data to accurately predict both sites of reactivity (SOR) and molecular reativity. Cross-validated predictions predicted with 89.8% AUC DNA SOR, and with 94.4% AUC protein SOR. The model also separated molecules electrophilically reactive with DNA and protein from nonreactive molecules with cross-validated AUCs of 78.7% and 79.8%, respectively. The model’s performances significantly outperformed reactivity indices from the quantum modeling literature, at both the site- and molecule-level. Moreover, we designed a selectivity score that predicts preferential reactions with the macromolecules, compared to common screening traps. Across the whole data set (2803 molecules), these scores summed to 257 (9.2%) and 227 (8.1%) molecules predicted to be, respectively, reactive only with DNA and protein, and therefore not detected by common screening assays. SOR of reactivity modeling across diverse chemicals is a pioneering approach that can be used to not only predict molecular reactivivity, but also suggest subtle structural modifications to minimize toxicity while retaining efficacy.

If you would like to use results from XenoSite Reactivity Predictor in your publications, please cite the following papers:

Hughes, T. B., Miller, G. P., Swamidass, S. J. (2015). Site of Reactivity Models Predict Molecular Reactivity of Diverse Chemicals with Glutathione. Chemical Research in Toxicology, 28(4), 797-809.

Hughes, T. B., Dang, N. L., Miller, G. P., Swamidass, S. J. (2016). Modeling Reactivity to Biological Macromolecules with a Deep Multitask Network. ACS Central Science, DOI: 10.1021/acscentsci.6b00162

XenoSite UGT

Uridine diphosphate glucunosyltransferases (UGTs) metabolize 15% of FDA approved drugs. Lead optimization efforts benefit from knowing how candidate drugs are metabolized by UGTs. The XenoSite UGT model predicts sites of UGT-mediated metabolism on drug-like molecules. In the training data, the sites of metabolism of 2839 UGT substrates are identified by our method with 86% (Top-1) and 97% (Top-2) accuracy. Both the size of this data set and our results improve over those of previously published UGT-metabolism prediction methods.

If you would like to use results from XenoSite UGT Predictor in your publications, please cite the following paper:

Dang, N. L, Hughes, T. B., Krishnamurthy, V., and Swamidass, S. J. (2016). A Simple Model Predicts UGT-Mediated Metabolism. Bioinformatics, DOI: 10.1093/bioinformatics/btw350

Acknowledgements

The authors would like to extend a special thanks to Brook Hauser for her help with web design and graphics.