Many applications of systems biology such as the simulation of isotope labeling experiments and pathway inference in metabolic engineering rely on atom-level representation of metabolism. Because of its challenging complexity and the difficulty of obtaining correct atom maps of biochemical reactions, the current understanding of the atomic level of metabolism cannot yet fully explain the outcome of stable-isotope labeling experiments. To our knowledge, iAM.NICE is the only computational tool for automatic mapping of single atoms in metabolic reactions, pathways and networks, that ensures the correctness of the mapping based on biochemical reaction mechanisms. iAM.NICE is an extension of BNICE.ch, which reconstructs biochemical reactions using expert curated, generalized biochemical reaction rules. However, inferring differential flux profiles from different atom-mapped metabolic pathways remains a challenge. In this study, we apply iAM.NICE to the core metabolism of E. coli to understand (i) the impact of different physiological states on the distribution of atom labels in the compounds of the central carbon metabolism, and (ii) the consecutive relation of labeled biomass precursor metabolites to the possible labeling distributions in the biomass building blocks (BBB). Our results show that different physiological states of a cell give rise to different distributions of carbon atoms in BBB precursor metabolites, information that is not only crucial for the design and the analysis of isotope labeling experiments, but also deepens our understanding of metabolism at the atomic level. Finally, our method can be easily extended to study in silico the conversion of elements other than carbon, and it can be applied to study organisms other than E. coli. The exhaustive results of our study can be consulted on our website (http://lcsb-databases.epfl.ch/pathways/LabelingList ). Access is freely available for academic purpose upon subscription.