Prediction of novel doping agent using an in silico model that integrates chemical, biological and phenotypic data

The anticipation of novel doping agents is usually not quick enough to prevent their misuse among athletes. Current techniques e.g. High performance Liquid Chromatography (HPLC) to detect novel doping agents usually involves a time­ consuming process. Computational method or in silica model offers a quick and accurate method to detect potential doping agents. The in silica model used in this study integrates chemical, biological and phenotypic data to identify novel doping agent. In this study, two different training sets, termed as biological and phenotypic were compiled and three molecular descriptors (MACCS, ECFP4, FCFP4) and two machine learning algorithms (Naive Bayes and Decision Tree) were employed to build the predictive models. These models were validated using five-fold cross validation to evaluate their predictive power. Both biological and phenotypic models were then combined using Joint Belief model. Sensitivity or the ability of the model to identify doping agent accurately between individual and combination models were compared. This study found that the combination model predicts better compared to the individual models with highest sensitivity of 0.962. Four compounds, Meldonium, Mitragynine, Pipradol and Synephrine were then tested in both individual and combination models. All four compounds were predicted as non-doping agents using biological models and were predicted as doping agents using phenotypic models. Combination of both biological and phenotypic models predicted all four compounds to be doping agents. The statement are supported by study done on Meldonium which is an anti-ischemic drug that was recently banned due to its effect on myocardial function that improves athlete's endurance. Mitragynine which is listed in WADA Monitoring Program, has also the potential as novel doping agent due to its binding activity on x-opioid receptors. The activation of x-opioid receptors produce analgesic effect which also aids in endurance of athletes. Hence, the use of in silica predictive model allowed the detection of novel doping agents quickly and can prevent the spreading of them among athletes.