Lithium is used (in the form of soluble salts) to treat bipolar disorder and has been considered as a possible drug in treating chronic neurodegenerative diseases such as Alzheimer's, Parkinson's, and Huntington's diseases. One of the proposed mechanisms of Li+ action involves a competition between the alien Li+ and native Mg2+ for metal-binding sites and subsequent inhibition of key enzymes involved in specific neurotransmission pathways, but not vital Mg2+ proteins in the cell. This raises the following intriguing questions: Why does Li+ replace Mg2+ only in enzymes involved in bipolar disorder, but not in Mg2+ proteins essential to cells? In general, what factors allow monovalent Li+ to displace divalent Mg2+ in proteins? Specifically, how do the composition, overall charge, and solvent exposure of the metal-binding site as well as a metal-bound phosphate affect the selectivity of Li+ over Mg2+? Among the many possible factors, we show that the competition between Mg2+ and Li+ depends on the net charge of the metal complex, which is determined by the numbers of metal cations and negatively charged ligands, as well as the relative solvent exposure of the metal cavity. The protein itself is found to select Mg2+ over the monovalent Li+ by providing a solvent-inaccessible Mg2+-binding site lined by negatively charged Asp/Glu, whereas the cell machinery was found to select Mg2+ among other competing divalent cations in the cellular fluids such as Ca2+ and Zn2+ by maintaining a high concentration ratio of Mg2+ to its biogenic competitor in various biological compartments. The calculations reveal why Li+ replaces Mg2+ only in enzymes that are known targets of Li+ therapy, but not in Mg2+ enzymes essential to cells, and also reveal features common to the former that differ from those in the latter proteins.