The formation of a complex between p21(ras) and GAP accelerates the GTPase reaction of p21(ras) and terminates the signal for cell proliferation. The understanding of this rate acceleration is important for the elucidation of the role of Ras mutants in tumor formation. In principle there are two main options for the origin of the effect of GAP. One is a direct electrostatic interaction between the residues of GAP and the transition state of the Ras-GAP complex and the other is a GAP-induced shift of the structure of Ras to a configuration that increases the stabilization of the transition state. This work examines the relative importance of these options by computer simulations of the catalytic effect of Ras. The simulations use the empirical valence bond (EVB) method to study the GTPase reaction along the alternative associative and dissociative paths. This approach reproduces the trend in the overall experimentally observed catalytic effect of GAP: the calculated effect is 7 +/- 3 kcal/mol as compared to the observed effect of approximately 6.6 kcal/mol. Furthermore, the calculated effect of mutating Arg789 to a nonpolar residue is 3-4 kcal/mol as compared to the observed effect of 4.5 kcal/mol for the Arg789Ala mutation. It is concluded, in agreement with previous proposals, that the effect of Arg789 is associated with its direct interaction with the transition state charge distribution. However, calculations that use the coordinates of Ras from the Ras-GAP complex (referred to here as Ras') reproduce a significant catalytic effect relative to the Ras coordinates. This indicates that part of the effect of GAP involves a stabilization of a catalytic configuration of Ras. This configuration increases the positive electrostatic potential on the beta-phosphate (relative to the corresponding situation in the free Ras). In other words, GAP stabilizes the GDP bound configuration of Ras relative to that of the GTP-bound conformation. The elusive oncogenic effect of mutating Gln61 is also explored. The calculated effect of such mutations in the Ras-GAP complex are found to be small, while the observed effect is very large (8.7 kcal/mol). Since the Ras is locked in its Ras-GAP configuration in our simulations, we conclude that the oncogenic effect of mutation of Gln61 is indirect and is associated most probably with the structural changes of Ras upon forming the Ras-GAP complex. In view of these and the results for the Ras' we conclude that GAP activates Ras by both direct electrostatic stabilization of the transition state and an indirect allosteric effect that stabilizes the GDP-bound form. The present study also explored the feasibility of the associative and dissociative mechanism in the GTPase reaction of Ras. It is concluded that the reaction is most likely to involve an associative mechanism.