Altered metabolism is a hallmark of cancer, but little is still known about its regulation. Here we measure transcriptomic, proteomic, phospho-proteomic and fluxomics data in a breast cancer cell-line across three different conditions. Integrating these multiomics data within a genome scale human metabolic model in combination with machine learning we systematically chart the different layers of metabolic regulation in breast cancer, predicting which enzymes and pathways are regulated at which level. We distinguish between two types of reactions, directly or indirectly regulated. Directly-regulated reactions include those whose flux is regulated by transcriptomic alterations (~890) or via proteomic or phospho-proteomics alterations (~140) in the enzymes catalyzing them. Indirectly regulated reactions are those that currently lack evidence for direct regulation in our measurements or predictions (~930). Remarkably, we find that the flux of indirectly regulated reactions is strongly coupled to the flux of the directly regulated ones, uncovering a hierarchical organization of breast cancer metabolism. Furthermore, the predicted indirectly regulated reactions are predominantly bi-directional. Taken together, this architecture may facilitate the formation of stochiometrically consistent flux distributions in response to the varying environmental conditions incurred by the tumor cells. The approach presented lays a conceptual and computational basis for a more complete mapping of metabolic regulation in different cancers with incoming additional data.