How far away is complete reaction simulation, where you can just enter molar ratios and solvent, temperature, etc. and have the computer spit out the expected reaction products? It seems possible but just not worth the massive computation time that would be required.
Related, what's the best way to predict the feasibility of reactions computationally? It seems like you should be able to compute rate constants or something to see if there's any appreciable chance the reaction will actually occur. I have used spartan before but i have no idea how to do anything but predict spectra and other static properties of single molecules.
Software like Gaussian, GAMESS, MOLCAS, Orca etc. can help you calculate many chemical properties of relatively small systems and QM/MM software can help you model larger groups.
You can plot the potential energy surface of all geometric configurations of a set of atoms and find yourself a path over it through stationary states like locak minima & transition states who are characterized by a single imaginary fequency obtained via the Hessian matrice.
But even then you do not have the complete reaction scheme of the system. You need to probe intersystem crossing probbailities etc with the Pauli-Breit hamiltonian, excited states... and you need to have the knowledge to operate all these functions at a sufficient level, like choosing the level of theory (CASSCF, DFT, Coupled Cluster, ...)
All in all, I would recommend Orca by Frank Neese since it was free the last timne I used it and very very capable
or write your own sotware but handling many center integrals properly can be a bitch