Inferring oncoenzymes in a genome-scale metabolic network for hepatocytes using bilevel optimization framework

Cancer cells exhibit unusual metabolic activity, characterized by high rates of glucose consumption and lactate production, even under aerobic conditions, as well as increased glutamine catabolism and amino acid metabolism. Based on the behaviors of metabolic reprogramming, we can infer oncogenesis from a genome-scale model of cancer cell metabolism. This study establishes a bilevel optimization formulation that integrates the genome-scale metabolic model of hepatocytes, the Warburg hypothesis, and LC/MS experimental data to detect multiple-hit enzyme deficiencies that induce metabolic reprogramming in hepatocytes. A nested hybrid differential evolution algorithm was employed to solve the bilevel optimization problem. The results predicted dopa decarboxylase (DDC) to be an influential enzyme that causes oncogenesis. A cluster analysis of the flux variations for different enzyme deficiencies obtained through a flux variability analysis showed that DDC is a dominant overexpressed enzyme, and it was classified into a group with similar trends of flux and metabolite alternations.