Quantitative, predictive modeling of biochemical networks: A machine learning and information-theoretic approach