We report results of detailed empirical valence bond simulations that model the effect of several amino acid substitutions on the thermodynamic (Delta G degrees) and kinetic activation (Delta G(double dagger)) barriers to deprotonation of dihydroxyacetone phosphate (DHAP) and D-glyceraldehyde 3-phosphate (GAP) bound to wild-type triosephosphate isomerase (TIM), as well as to the K12G, E97A, E97D, E97Q, K12G/E97A, 1170A, L230A, 1170A/L230A, and P166A variants of this enzyme. The EVB simulations model the observed effect of the P166A mutation on protein structure. The E97A, E97Q, and E97D mutations of the conserved E97 side chain result in <= 1.0 kcal mol(-1) decreases in the activation barrier for substrate deprotonation. The agreement between experimental and computed activation barriers is within +/- 1 kcal mol(-1), with a strong linear correlation between Delta G(double dagger) and Delta G degrees for all 11 variants, with slopes beta = 0.73 (R-2 = 0.994) and beta = 0.74 (R-2 = 0.995) for the deprotonation of DHAP and GAP, respectively. These Bronsted-type correlations show that the amino acid side chains examined in this study function to reduce the standard-state Gibbs free energy of reaction for deprotonation of the weak alpha-carbonyl carbon acid substrate to form the enediolate phosphate reaction intermediate. TIM utilizes the cationic side chain of K12 to provide direct electrostatic stabilization of the enolate oxyanion, and the nonpolar side chains of P166, 1170, and L230 are utilized for the construction of an active-site cavity that provides optimal stabilization of the enediolate phosphate intermediate relative to the carbon acid substrate.