Skip to main content

Patents and IP

How Do You Find A New Compound to Patent?

I recently had a question from an interested reader, who’s outside of medicinal chemistry, wondering about how we find room to work in patent space. From his viewpoint, there seemed to be many areas that were so worked over that he found it had to see how any new project could make headway, or be able to keep track of what had already been claimed. There’s certainly something to that. Some structural classes have been beaten on pretty thoroughly over the years, for sure, but there are ways.

A quick bit of background, keeping in mind that I Am Not A Patent Lawyer. I’m going to focus on “composition of matter” claims, where you claim the actual chemical compound itself, rather than method/use claims, which are a different beast. Composition of matter is actually one of the fundamental pieces of US patent law itself. Title 35 of the US Code, describing what’s eligible for a patent, says that you can get one for “any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof”. So for a drug, that’s an exact description of the new molecule you’ve made.

You’ll note that it has to be both new and useful (novelty and utility), but for a drug patent, utility can be a pretty easy bar to clear – this new compound is useful for giving to patients in need of a treatment for medical condition X. (That’s not always the case, but we’ll continue with the ones where it is). And that takes us to “new”, which is where the original question comes in. After there’s been a lot of patent activity in a particular chemical class, just how new can a compound be?

Once in a while, you’ll see a patent that claims one compound and one compound only. But that’s rare; most of the time there’s a structure with a bunch of R and X and Y groups hanging off of it. That’s a “Markush” structure, named after Eugene Markush, who filed a patent in 1924 whose prosecution ended up establishing this as a legitimate way to make claims in chemical space. You can vary things by length of a chain, by position and place of attachment, by listing acceptable substituents, and so on, and as you’d imagine, this can rapidly lead to a huge universe of possible compounds. I’ve seen some truly insane Markush structures over the years, things that if fully enumerated would reach into hundreds of millions of compounds.

If you have a hot new molecule you’re interested in developing, you’ll want to search these things. Markush searching is painful – no other way to describe it. The existing computational methods for doing it invariably pull in far more patent filings than you really have to worry about, because of all the vague claim language out there, and you have to dig through them manually at that point. But you can sort things down a bit. For example, you’d first want to see if any of those patents are going after the same sort of activity/target/disease that you are, because those are clearly the ones that you should be most worried about.

Then you have to get a sense of just how deep into the claims any overlaps seem to go. What you have to keep in mind is that these Markush lists are never fully exemplified – that is, the specific compounds are not all described with their unique physical properties to show that they were really prepared. There’s not enough time or money in the world for it. So you need to see what the patent in question really starts to zoom down on, what it “teaches toward”. A patent is required to state the best “embodiment” of the invention, and its claim are generally written in an order that narrows down on that. If your compound hits the outer penumbra of their Markush structures, but their patent then heads off in another direction when the claims start to get more serious and directed, that’s a good thing. At the other end of the scale, if they specifically exemplify your exact compound, you have a real problem.

It’s that latter situation that comes up in the heavily-worked-over parts of chemical structure space. Not only are there a heap of Markush overlaps, but there are an awful lot of real compounds, too, and once they’ve been disclosed, really disclosed as in “here’s how you make this one and here’s its unique spectral data to prove that we did it”, you absolutely cannot get a composition of matter patent – it is “prior art” and no way around it. For the most part, you just want to stay away from areas like this – surely you have another structural series that’s just as interesting? If you have to pick your way through the cratered landscape, though, it can often be done. Chemical space is large, insanely large, and there are almost always variations that will allow you to prepare a compound that has never been described. Fluorinate it over here, add a heteroatom over there, tie a ring back – there’s usually something, and it all depends on how much you want that part of chemical space and how much time and money you’re prepared to spend to find the right piece of it. It would definitely serve you best if some of these changes also made a difference in the compound’s activity and usefulness, since there’s a “doctrine of equivalents” that basically says that a difference that makes no difference is no difference, but you can generally find enough of a ledge to stand on.

The general feeling among med-chemists is that any patent can be broken by enough of a determined effort, but it’s a bit like putting a lock on a door. Any lock can be defeated by enough of a determined effort, too – you just have to figure out how strong to make it (that is, how much money to spend on it) to protect what’s behind it. Shoring up a patent claim involve exemplifying more compounds in as many plausible variations as you can think of, until someone cries “Halt! Enough!” The hope is that you haven’t left any incredible holy-cow analogs untouched, because those would have the advantage of not being obvious “to one skilled in the art” as the patent lawyers say, and you certainly weren’t teaching toward them in your claims, either. Every drug composition-of-matter patent is, in theory, vulnerable to this sort of thing, but for the most part, given the odds of this happening, it’s a better use of time and money to try to find something that’s more in the clear, unless the patent in question leaves some obvious gaps and the potential benefit is clear.

31 comments on “How Do You Find A New Compound to Patent?”

  1. Small Biotech says:

    Derek, does the data in this article regarding compounds exemplified per patent jive with your personal experience and perception of state of the art?

  2. Anon says:

    Technically, there are no novel compounds because the generic R group sticking to an old patented core structure could conceivably contain your new and inventive core.

    Therefore, “novelty” is interpreted strictly according to whether the specific compound has previously been described in exact detail, with no ambiguities in generic R groups, etc.

    Most of the time, the question of patentability focuses on “obviousness”, according to whether somebody “skilled in the art” would consider what you claim as an obvious extension of what has been described before. This is of course highly subjective, which is why you need a good patent lawyer to argue your case because most patent examiners will say that everything is obvious based on some loosely relevant references.

    1. B says:

      Hi Anon

      “Technically, there are no novel compounds because the generic R group sticking to an old patented core structure could conceivably contain your new and inventive core.”

      This is not the case. R groups are defined in patents, even broadly by terms such as ‘alkyl’, ‘aryl’ etc. This is considered a disclosure of a particular compound. However, because it is considered a disclosure of a particular compound, it also means that it is absolutely possible to find novel chemical space.

      Don’t forget, it’s not just new in terms of patent disclosures. A compound disclosed in a journal article will also destroy novelty.

      As Derek highlights, it is also possible to still claim a compound per se as being ‘novel’, even if it is disclosed in a generic R=alkyl-type structure of a patent (you can claim these as a “selection” patent).

      It’s not all doom and gloom! 🙂

  3. G says:

    This goes back about 30 years. We had a series of compounds where certain imidazole substituents gave terrific, and highly specific, activity. There were patents claiming the identical skeleton, with various amine substituents at the same position as our imidazoles. However, the compounds with “ordinary” amines were totally inactive, per a publication by the patent owner. We were able to patent our compounds on the basis of “non-obviousness,” as the prior art suggested that the imidazoles would not be active. “Dancing between the raindrops” was the way our management described it.

  4. Polynices says:

    How plausible do you think it is that useful drugs (or even just real but not very useful) drugs are “hiding” in all those old patents — not characterized (or properly screened for) by the patent holder but fending off anyone else from looking closely enough? Or is this not really how it works out in practice? Genuinely curious what real chemists think — I have just enough background for the question to occur to me, no clue how to answer it properly.

    1. LLCoolChris says:

      I had the same question, Polynices. Along those lines, is it common to arrange a deal for use of a compound named in a patent for usage in another therapeutic area?

      1. Simon says:

        I don’t know how common but it’s definitely done. I tried to do it myself a couple weeks ago but didn’t see worth it. Celgene licensed thalidomide analogs from EntreMed in 2003 and seems to have done quite well with them!

    2. John Wayne says:

      In my opinion, the hard step is drug discovery is deciding how to look to new drugs. Ignoring potential leads you cannot get composition of matter claims on is a flea fart in the face of our ignorance of biology.

    3. It definitely happens in practice. A good example taught to all med chemists at Pfizer was the case of fluconazole. ICI held a patent in which fluconazole was covered by the Markush but was not specifically exemplified. Pfizer made fluconazole and found that it had some distinct advantages versus the closest examples actually made and described in the ICI patent: lack of teratogenicity and better solubility (vs the 2,4-dichloro analog) and better antifungal potency (vs the 4-fluoro analog).

      These advantages allowed Pfizer to meet the criteria for obtaining a patent of selection (ie selection of a specific, unexemplified subset of a previously claimed Markush). This meant that there were now two valid patents protecting fluconazole. It’s important to remember that patents confer negative rights (ie the right to exclude others from exploiting an invention) rather than positive rights (ie the right to exploit one’s own invention), so that left Pfizer and ICI in the interesting position of each being able to prevent the other from selling fluconazole. Happily, ICI agreed to grant Pfizer a license, allowing fluconazole to be commercialized.

      A brief summary of the situation is here:

    4. Patent troll says:

      That’s a good question. I think the answer is that there probably are a lot of “real” but not very useful drugs hiding in old patents. To explain – the way things usually work out is that the development team eventually has to pick one horse (a compound) to ride into the clinic. In the unlikely event the horse crosses the finish line (and it’s a big indication), there are usually several me-too compounds from other companies on the market within a few years. In that scenario, there probably are a bunch of decent drug molecules in the original patent that no one will ever develop because the market is too crowded (i.e. real but not useful). Although ironically, as in the fluconazole case, the me-too drugs are sometimes drawn from unexemplified compounds that fall under the original Markush.

      Most of the time the horse doesn’t make it, and everyone is left wondering “well, was it just a crappy compound or is the whole mechanism bogus?” There usually isn’t a clear answer to this and consequently the whole program is scrapped. I suspect there are quite a few cases (especially at smaller companies with less development experience) where the wrong molecule was advanced and they inadvertently burned off an entire therapeutic approach. In those instances, there may well be real and useful drugs in the old patents. But good luck convincing anyone to part with enough cash to find out.

      Also, patents expire after 20 years, and realistically you only get one kick at the can to develop a molecule from each patent – if the patent only has a few years left, that usually isn’t enough time to make your money back even if the drug works. So again, even if the molecule looks promising, if the patent is too long in the tooth no one will pay to develop it.

  5. “hundreds of millions of compounds”

    How quaint 8-). The infamous Canadian Viagra ‘446 patent had approximately 260 quintillion structures enumerated by its Markush structures. Granted, that’s one of the reasons that the SJC ruled that the patent was overly broad in the first place, but still… 😎

  6. PharmaJohn says:

    Successful patent busting really depends on how well you know your chemical space and target space. Other considerations are: expired patent claims, surrendered patents, and remaining patent years. All these are well beyond the realm of most medchemists. Companies and careers are build around patent assets, and yet there aren’t enough discussions about patent blocking and defending strategies. Dancing between the raindrops indeed.

  7. JAB says:

    Another way to fence off chemical space that is totally legitimate is to write your patent however broadly as you can with the examples you have, and then publish compounds outside or inside of that fence in journals- the compounds then cannot be patented by others since the prior art is there. In some cases you might file additional patents IF there is a major unexpected advantage to your new compounds, say oral bioavailability or better PK.

  8. Anon says:

    On a related note, I was wondering if there are any examples of drugs which have been effectively protected by trade secret only, i.e., without publishing a patent?

    1. CB says:

      Trade secret protection wouldn’t really work for pharmaceuticals. To be considered a trade secret, the compound has to derive value from not being generally known. Once you sell the compound, it’s no longer a trade secret. You could certainly keep the process for making a drug as a trade secret, but anyone skilled in the art could likely arrive at your process (or a reasonable facsimile thereof) using basic retrosynthetic analysis.

      1. Anon says:

        I was thinking like Coca Cola’s secret formula, where they sell the substance but still manage to keep it secret. Or have people figured that out already from chemical analysis?

      2. Anon says:

        … and in any case, why don’t companies wait until just before launch to apply for a patent? Surely the company and FDA should be able to keep the formula secret until then, no?

        That would give an extry 10 years or so of market exclusivity at peak sales!

        1. Isidore says:

          The problem is that you cannot be sure that someone else will not file for a patent on the same compound which you are secretly developing. Since the first to file will be granted the patent, it would be disastrous to find this out after having spent a ton of money developing your blockbuster drug.

        2. Lyle Langley says:


          Ideally that would be great, however, you run the risk of being scooped if you wait too long. As the project progresses (and if it’s interesting) people will start to figure out the structure – nothing can be kept a secret. It’s not first to invent any more, it’s first to patent.

        3. Mad Scientist says:

          … and in any case, why don’t companies wait until just before launch to apply for a patent? Surely the company and FDA should be able to keep the formula secret until then, no

          That’s an outstanding way to possibly flush hundreds of millions of dollars down the toilet, and / or lose a potential blockbuster. In today’s tighter IP space, with many groups working on similar conceptual therapeutics, you would want to resolve the patent issues at an early stage before committing your resources for advanced drug development. It’ll be like Christmas for the patent owner!

          Think about it – your great magical pill is ready to launch, but someone else was granted the IP rights while you were busy finishing up clinical studies on your secret compound. In fact, the Regulatory agencies will almost invariably demand submission of legitimate patent information along with the IND before embarking on human studies.

          1. overthetop says:

            Plus, there is no guarantee that your last minute patent application will be granted. Many patent applications are abandoned every year because they are simply not novel or they are obvious. Appealing those negative outcomes can take years to resolve, so you wouldn’t know if your wonder drug will have any IP protection until millions of dollars have already been invested.

          2. steve says:

            I hate to break it to you guys but Anon is right. That is indeed the strategy most large pharma take – they don’t patent until late in the game in order to maximize patent life.

    2. exGlaxoid says:

      Most over-the-counter or non-prescription drugs are protected only by Ttrademarks, and some make a fortune. Look at Tylenol, Sudafed, Neosporin, Pepcid, and 1000’s more. They are where the money is, as you can make them for very little money in old factories, market them some every year, and many brand name OTCs make $100 million to $1 billion a year. For a while, big pharma companies were shedding OTCs, now they are back to buying them up, since they are much less regulated and there is no insurance company to argue with. Many brands have been cash cows for years, look at Bayer aspirin for example.

      1. Lyle Langley says:

        You are correct, big Pharma “tries” to patent as late as possible; however, that is not always possible – if “competitive intelligence” states the area is quite competitive, then they still patent early. And, as stated above, patenting late (and when you say late, it’s not as late as the original poster) has its own issues as well.

  9. Moses says:

    For Anon, 5:08.

    Companies patent as soon as possible to prevent another company from getting there first.
    There was a case in the 90s where two companies (I think one was Cyanamid) applied to patent the same structure for a fungicide, with one being applied for only a day before the second.

  10. petros says:

    and to emphasise the point made by Moses I remember seeing the same compounds claimed and exemplified in PDE4 inhibitors from Celltech and RPR that were published 1 week apart.

  11. Lyle Langley says:


    Back in the 90’s it wouldn’t have been a case of who applied for the patent first (at least not in the US). It was first to invent then and they would be looking at that and not the filing date – which obviously gets much messier.

  12. steve says:

    As I mentioned in my reply to Anon, most large pharma don’t rush to patent. It used to be that you could “submarine” your patents so you could submit early and it wouldn’t be published for a while. Now many (if not most) will patent much later in the game in order to maximize patent lifespan. In general, pharma won’t touch compounds without composition of matter because it’s too easy to get around methods claims. Another strategy that can be used is to wait until a drug is close to market and then patent the metabolites. Then, even if the original patent expires, generic companies are precluded from market entry because they’ll violate the patent on the metabolites. Figuring out the timing of patent submission vs market entry is not a simple matter.

  13. newnickname says:

    One place to look for “new” compounds is “old” literature. Anything published prior to the 1960s has no NMR (I’m referring to routine organic NMR, Varian A60, 1961, not 1940s-1950s cutting edge research NMR); prior to the late 1950s, no routine IR; prior to the 1940s, no routine UV. Examples from the older literature describe compounds that would be unreasonable to expect from the procedures described. Using contemporary methods of synthesis and structure proof, some of these “old” incorrect compounds might actually become “new” new compounds. I’ve got a few favorites that, the last time I checked, still did not exist in authenticated post-1970s literature but could probably be made by modern rational sythesis.

    There are some “easy” publications to be had from that, too. Repeat the old reactions and figure out what they really synthesized using Xray, NMR, etc..

  14. steve says:

    My question is about repurposing drugs. Given that pharma wants composition of matter claims, how does it deal with repurposed drugs where there are only methods claims?

  15. Dylan says:

    “I’ve seen some truly insane Markush structures over the years, things that if fully enumerated would reach into hundreds of millions of compounds.”

    FYI- if one of those millions of compounds turns out to be inactive, the entire Markush family of compounds is invalid. The patent may fall. Which is why it’s nice to have some individual claims for the compounds you know are active.

    But what that Markush diagram does do is bring millions of compounds into the ‘prior art’. Hence anyone who wants to work (or is working) on the compounds is screwed. There is no incentive to invalidate the Markush structures in the first place if you can’t come out with an alternative. It’s a good defensive strategy that destroys incentive.

    Markush structures should be banned. But you really don’t need patents for drugs, all you need is FDA approval.

Comments are closed.