One of the many interesting points that came up in the comments to my Brazil posts was from a reader who wondered how easy it was for someone to get the structure (and synthetic route) for a marketed medicine. In other words, what are the barriers to ripping off someone's patent?
And as some of my industry readers correctly told him, those barriers are few indeed. The structure of an FDA-approved drug is a matter of public record, and is included in the required package insert with every prescription. And the patents are required to tell you how to make it. In fact, in almost every case, the only reason that a patented drug isn't copied is patent law itself. That gets right to the concept of what a patent is: in return for a defined period of monopoly, the patent holder agrees to disclose all the relevant facts about the invention.
For a drug industry patent, that means that the generic structure of the claimed compounds is in there, as well as at least the names and some physical properties of all the specifically exemplified compounds. (Chemical names are systematic enough to determine the structure, but most patents have the structural drawings, anyway.) General chemical routes are laid out, along with specific examples that get right down to the bench-level details. A chemical matter patent, in order to be valid, has to teach a reader of ordinary skill in the art how to make the claimed compounds. If it doesn't, it shouldn't have been granted, and can be invalidated. (This isn't merely an academic point - part of the reason that Bayer got ciprofloxacin (Cipro) from out of a SmithKline patent claim was that the routes shown couldn't make the compound.)
Now, there's still a bit of wiggle room. You don't have to list biological activities of all the compounds that you exemplify, although it does build a stronger case. If you're claiming a particular biological activity, you have to show how you assayed for it, but often you'll see the detailed assay procedure along with a note to the effect that "the compounds of this invention were tested according to this protocol and showed inhibition constants of between 0.01 and 10 micromolar."
That satisfies the legal requirements, but it doesn't satisfy the reader very much, which is the whole idea. There's a constant guessing game when your competitors aren't on the market yet and haven't disclosed much about their clinical trials. Ten nanomolar to 10 micromolar is a wide range of activity, from the very interesting to the surely useless, and there's often no way of finding out which compounds were the best.
You can narrow things down by some detective work, though. Looking over the specific examples in the patent(s) can tell you what chemical classes they spent the most time on, and what regions of the molecule seemed to be most important. If your competitor has gone to the trouble of filing a process patent (covering a specific route to make certain compounds, as well as the compounds themselves), then you can be pretty sure that they're interested in that class. And looking at the specific examples in the patents will sometimes show an interesting pattern: ten, twenty, fifty compounds are all shown being prepared in 20-milligram amounts, followed by one that's made on a ten-gram scale. Aha!
Now, in the absence of a process patent, you're still not guaranteed that the method shown in the original patent is the one that they're using to make it on large scale. Actually, you're more like guaranteed that it's not. Ten grams is one thing, but a thousand-kilo route is quite another. A patent is required to disclose the best mode to realize the invention that you know of at the time of filing, and that's generally before the process chemists have had time to work on the synthetic route. (Patent cases have hinged on the timing of those steps.)
But a good process group can usually take the structure and the patent route and work something out, for a tiny fraction of the time and effort it took to find the compound the first time and get it tested and approved. And if it weren't illegal, that's exactly what people would do. And then who would go to the trouble of being first?