The catalytic power of enzymes reflects billions of years of refinement by evolution. Thus, understanding enzyme action should be very instructive as a guide for general principles of catalysis. This review describes the emerging use of computer modeling as a key toll in correlating enzyme structures and catalysis. The review starts by describing general computer simulation approaches for calculations of free energies and related properties. It then moves to methods that allow one to evaluate the free energy of enzymatic reactions. This includes hybrid quantum mechanical/molecular mechanics (QM/MM) molecular orbitals methods, the empirical valence bond (EVB) method, and other approaches. Emphasis is placed on specifying the requirement from reliable simulation approaches, including the need for accurate potential surfaces and a proper configurational sampling. The ability of different approaches to satisfy these serious requirements is critically examined. Next, we examine what has been learned from simulations of enzymatic reactions about the origin of enzyme catalysis. It is pointed out that consistent simulation studies concluded that electrostatic effects are the key catalytic factor and that the corresponding catalytic energy is stored in preorganized polar active sites. Studies that examined the magnitude of other catalytic contributions and the validity of the corresponding catalytic proposals are also considered. These include studies of various ground state destabilization effects (e.g. entropic, strain, and desolvation contributions), studies of dynamical effects, and studies of the low barrier hydrogen bond proposal. It is concluded that these contributions do not play a major role in enzyme catalysis. Finally, the future of computer simulations as a critical tool in studies of enzymatic reaction is briefly considered.