The sorts of data I would like to find are things like how it changes based on temperature (melting, boiling, ionization, ect), hardness of the resulting material, malleability, density, ect.
I could be wrong but large-scale properties like hardness and malleability (especially malleability) might be extremely difficult, maybe even impossible, to predict without an insane attention to detail. For example, the hardness and malleability of smelted steel will depend heavily on how it was smelted. And it would be difficult to turn the smelting process into a set of parameters to input into a simulation or model.
Things like melting, boiling and ionization are easier to estimate, though (I think). It also depends on the level of complexity of interactions you want to deal with, though. First, the interactions of complex molecules are exponentially more difficult to figure out than interactions of simple molecules (and last I checked we're pretty bad at computing interactions of complex molecules without excessive babysitting and fine-tuning). Also if you care about catalytic reactions, you're in a whole new situation.
I remember you bringing this up in another thread in the context of a future production system for games; I'm skeptical that we'll see powerful enough commercial computers to do fast enough chemical simulations to be practical in a game (people would want to try out lots of combinations, and they wouldn't be happy with waiting). The day we see powerful enough computers and good enough software for that is the day that materials science becomes near obsolete, except for the more esoteric things; after all if any schmuck can predict what you're looking for in a negligible time-frame, who needs full time scientists to do it?
