Just to expand on the existing answer and make it a bit more clear, multitrack audio doesn't really work that was (sorry). Once a file is mixed down to stereo for distribution, there is no way to split it back into it's component tracks.
The only real option (aside from a midi arrangement like other mention) is to get the "session files", which is all of the original multitrack recordings along with the program data from the application used to record them.
There are some examples of this available, but you probably won't have a great selection to choose from, but you'll have something to build a prototype from at least. Later you might consider commissioning your own recordings to get the session files or asking someone to use theirs.
One example (I haven't used these, so I can't vouch for quality): http://www.cambridge-mt.com/ms-mtk.htm
An analogy that might help with the concept:
If you "flatten" a file in photoshop and then export to a JPEG, you won't be able to get split the JPEG back into the different layers used to make it.
More abstractly: When cooking a meal, if you mix various spices and ingredients to make a sauce, you can't really get those components back after the fact.