Metabolism of xenobiotics plays a crucial role in safety and efficacy of not only drugs, but also cosmetics or agrochemicals. However, even despite significant advances in analytical methods, experimental determination of metabolites still remains a labour-intensive and costly endeavor. Therefore, the prospect of an accurate and fast computational method for prediction of human metabolism is very attractive. FAME 3 represents a humble step in this direction by providing predictions of the so called sites of metabolism (SoMs). SoMs are atoms where a reaction catalyzed by a metabolic enzyme is initiated and, thus, their identification is useful in determining the metabolic fate of a compound. It can also serve as the first step in the direct prediction of formed metabolites (see GLORY).
FAME 3 is the third generation of the FAst MEtabolizer program developed for the prediction of SoMs. FAME 2 featured an improved machine learning methodology, but was only limited to a handful of cytochrome P450 (CYPs) isoforms.
FAME 3 uses the improved methodology of FAME 2 and applies it to a new data set that extends not only beyond CYPs in phase I transformations, but also includes data on many phase II reactions as well. The FAME 3 program uses a set of extra trees models that are trained on the MetaQSAR database, which is a large expert-curated database of metabolic transformations relating to both phase I and phase II metabolic enzymes. With data on more than 5,000 phase I SoMs and 1,200 phase II SoMs, this is a significant enhancement when compared to FAME 2.
In addition to FAME 2, FAME 3 also features an applicability domain measure (called FAMEscore), which evaluates how similar the environment of an atom is to the environments in the training set of the model. FAMEscore correlates with model performance and, thus, informs the program user about the likely accuracy of each prediction.
Enter SMILES, draw a molecule, or upload a file (.smi or .sdf) containing up to 1000 molecules. Click submit to start the calculation. You will then be forwarded to the result page.
The FAME 3 output for each submitted substrate is displayed on the results page after the calculation finishes (for about 50 molecules of average size, this should be within a few minutes). You can use the accordion buttons to display the prediction page for each submitted molecule one by one. The prediction page contains the structure of the substrate and a table with the predictions.
Each atom is assigned a unique ID within the molecule (the "Atom column"), which is also displayed at the corresponding atom position in the structure upon hovering over it with a mouse. The table is sorted according to the SoM probability assigned by the model (the "Probability" column) and the SoMs past the decision threshold of the model are also highlighted with yellow color. The FAMEscore value is shown in the last column of the table. Usually, FAMEscore values higher than 0.6 indicate that the prediction should be well supported by the model, but in some cases the model was found to perform well even if FAMEscore was lower (see the FAME 3 paper for further discussion).
Šícho, M.; Stork, C.; Mazzolari, A.; de Bruyn Kops, C.; Pedretti, A.; Testa, B.; Vistoli, G.; Svozil, D.; Kirchmair, J. FAME 3: Predicting the Sites of Metabolism in Synthetic Compounds and Natural Products for Phase 1 and Phase 2 Metabolic Enzymes. J. Chem. Inf. Model. 2019.
doi: 10.1021/acs.jcim.9b00376
Stork, C.; Embruch, G.; Šícho, M.; de Bruyn Kops, C.; Chen, Y.;
Svozil, D.; Kirchmair, J. NERDD: a web portal providing access to in
silico tools for drug discovery. Bioinformatics
2020.
doi: 10.1093/bioinformatics/btz695
To report a problem or give feedback, please go to the Feedback & Support page or contact us directly: