[Woopy]

I was wondering if chemdraw can actually figure out the products of reactions, and if not, is there any program that can do that?

[curiouscat]

I was wondering if chemdraw can actually figure out the products of reactions, and if not, is there any program that can do that?

Nope. It cannot.

And I do not know of any general program like that.

[IsotopeBill]

I know there have been attempts to develop such programs, but I don't believe they've been successful on any real scale.

There are programs such as SciFinder and Beilstein's Crossfire that basically mine the chemical literature.  Among other things, you can input your reagents, and see if anyone has reported the product of that reaction, or input a product to look for reported ways to make it.  The program is only giving you information that has been reported, however:  it is not "figuring out" the chemistry.

They are very very expensive, but if you're at a university, they may have a site license to the program or to a scaled-down educational version.  I recommend inquiring about them and, if they're available, seeing if you can get some time learning how to use them

[Woopy]

Well if anyone successfully makes a program like I described, they will surely be multimillionaires

[fledarmus]

That was a hot topic of research back when I was in grad school - 30 years ago. As far as I know, it still hasn't been carried out successfully.

[Woopy]

Why is it so hard to program something like this that it hasn't been done yet?

[fledarmus]

First, because chemistry isn't particularly predictive. There are a large number of different reactions possible for most compounds, all leading to different products, and very small changes in reaction conditions can lead to large changes in the outcomes. Then again, you have to consider the potential reactivity of all the products when exposed to the same reaction conditions. You have to specifiy the starting conditions thoroughly to restrict the problem to a small number of outcomes, and with that type of restriction, if the reaction hasn't already been run, predictions of products usually aren't very useful.

I am aware of two types of solutions that have been studied. One is essentially a database miner - you enter your starting material and a limited set of conditions (other reagents, temperature, etc.) and the program tells you the products that have been attained under those conditions. Reasonable access to massive reaction databases like Reaccs and SciFinder have made this approach obsolete. Why ask a program what sort of reactions have been found for a compound rather than simply searching the reactions for yourself?

The second is an expert system approach - you enter your starting material, tell it what type of reaction you want it to undergo (substitution, elimination, reduction, elimination, etc.) and it tells you the products you can expect. There is some utility for these if they can incorporate enough geometry to tell you when your system is overconstrained to undergo reaction or if certain places are unavailable as reaction sites, or even determining which of a number of possible reaction sites would be most likely to react under different circumstances. For the most part, though, they are only useful for people taking organic chemistry exams. You still have to come up with the conditions and run the reactions to see if they work, and again, you can get the same information from SciFinder or Reaccs by choosing your substructures properly.

The second reason is that this really isn't the way most people approach chemistry. Chemical research is usually directed to a targeted compound and finding reactions that will lead to the compound, rather than trying to predict the ways a compound will react to other reagents. You already have a pretty good idea of how the compound would react to those reagents, or you wouldn't have been looking at that compound to begin with. Maybe that was a little obscure? See if this helps - you are asking about a program that would allow you to input a compound A and some reaction conditions B and predict what compounds C would be formed. Usually in research, however, you are more interested in designing a compound C and trying to determine what starting materials A and reaction conditions B would be reguired to generate that compound. Very rarely is there call for taking a compound A, trying a bunch of reaction conditions B on it, and seeing what sorts of final compounds can be formed - you almost always know what final compound you want, and the work is in trying to find the reaction conditions and starting materials to make it as efficiently as possible.

[Arkcon]

Why is it so hard to program something like this that it hasn't been done yet?

To summarize what fledarmus: said, the topic came up in Chemistry 101 and basic Organic Chemistry, and the teacher responded, because no one wants such a programs, except students too lazy to learn the fundamental chemistry rules.  That comes off a little trite, too perfect an answer for the student.  But as fledermus: expanded -- there are two many conflicting situations to model them all.  Once you've restricted the rule set a little bit, you've practically solved it yourself, but until you do, you can make no useful predictions.