FightingDrugFailure was a multi-institutional project, where our role was to combine machine learning for optimization of metabolic models. Given the computational model of metabolism, our idea was to record its outputs under some selection of model parameters, and then study the role of these parameters using machine learning approaches. Our premise was that we can identify principal parameters, and parameters (pathways) that could be excluded from the model since they do not have significant effect on the observed outcomes.