Skip to main content

Drug Assays

A Magic Methyl, Spotted in the Wild

You hear medicinal chemists talking about the “magic methyl”, the big effect that a single CH3 group can have on potency or selectivity. Here’s a new J. Med. Chem. paper that shows one in action.That structure looks like a kinase inhibitor if anything ever did, and so it is. But small changes to it can make a big difference. As you see from those assay numbers, adding in one (R) methyl group hurts MERTK activity by six-fold while making no real change against TYRO3. But AXL activity goes down eight-fold, and FLT3 activity by something like 80-fold.

The effect is seen with a number of different variations around this core, and it’s obviously because there’s something that FLT3, in particular doesn’t like about having a methyl there. Specifically, it’s a methionine in MERTK that’s a leucine in FLT3, and the branched chain of the leucine just doesn’t leave enough room for a methyl in the ligand. As you’d expect, if you put in a methyl of the opposite stereochemistry, there’s a big effect as well – that (S) methyl compound is 22, 560, 13, and 543 nM on the four enzymes, so you can see that there’s an even more severe clash in all of them except TYRO3, which doesn’t care one bit.

I bring this up as an illustration of what med-chem is like, pretty much all the time. There will be parts of your molecule that are pretty insensitive – you can hang all sorts of crap off them, and if the project goes on long enough, someone will. Those, as you’d imagine, are generally the parts sticking out into solvent, which is where you’ll see people desperately sticking on methoxyethoxy chains or whatever to try to make the compound less of a brick in its solubility and pharmacokinetics. And there will be parts that you really can’t touch at all. I’m sure that if you start yanking those nitrogen atoms out of the structure above in favor of good ol’ carbons that it’ll start losing activity with great speed and thoroughness, for example.

It’s the in-between parts where we make our livings. That’s where you exploit differences in amino acid side chains in the protein binding sites, where you tickle bound water molecules or pick up pi-interactions. And these effects have always been a tricky thing to handle in modeling the activities that result, and they’re going to be a tricky thing for machine learning, too. “Activity cliffs” are a feature of most structure-activity relationships, and sometimes you don’t know that they’re there until you walk off of them (and sadly, we have no Wile E. Coyote physics to save us). Abrupt nonlinearity is not an easy thing to work into a model – if you don’t watch it closely, said model might lose its silicon mind a little bit trying to fit that stuff into a coherent picture. That’s because there may not be a very coherent picture: most projects end up with something like “This works, but not too much of it, and only if you have that thing over there, but if you have something in that other spot it kind of cancels that out, well, except if you have this other thing, but when you. . .”

And that’s just trying to get a handle on the primary potency numbers. When you add in cell membrane permeability, pharmacokinetics, metabolic stability, and all the rest of it, well, things can get pretty wooly. For instance, the authors of the paper above ended up switching that piperazine to a homopiperazine to get better half-life numbers in their rodent dosing. Some machine help in all this would be welcome, but I’m not sure when we’ll see that arrive. The data sets are not always large enough to be useful for machine learning, particularly when doing a multifactorial optimization like this, and they’re definitely not large enough earlier in a project. And we don’t have the knowledge to just model everything de novo, either. If you want to know how selective your kinase inhibitor is against the other kinases, you don’t sit down and model them all – you send them out to a screening panel to get the actual numbers. And if you’d like to know what other targets your compound might hit, well, good luck. There are broader screening panels, of course, that will run your compound through dozens of assays. But there are a lot more targets out there that we know little or nothing about, and the only way you’ll know if one of those is doing something off-kilter is to dose rats. Or, God help you, humans.

So what I’m saying is that medicinal chemists, for all the talk about machine learning and AI, are not going to be replaced in this part of their job – the main part, mind you – any time soon. None of us, machines or humans, are smart enough for that, or not yet.

23 comments on “A Magic Methyl, Spotted in the Wild”

  1. Wavefunction says:

    I agree that modeling and ML will almost always have trouble with activity cliffs, but the question is regarding how common these are. If the assumption that similar compounds bind similarly holds up enough number of times (an assumption that medicinal chemists do operate based on), then those techniques will simply have to be statistically good enough to make a tangible impact.

    1. t says:

      How does one define “similar”?

      1. Istvan Ujvary says:

        I know it when I see it

    2. Peter Kenny says:

      Hi Ash, perhaps the issue for ML models in this context is not so much that the models have trouble with activity cliffs but that a ML model that has not seen the activity cliff is being extrapolated out of its applicability domain when you use it to make predictions for the magically methylated analog.

      1. Wavefunction says:

        Good point about the applicability domain Pete. But even if the ML model had seen the activity cliff it would rightly assign it a low probability because it’s not that common. That’s why I am a fan of ML models giving me results with probabilistic scenarios, a bit similar to how climate models give you different scenarios with different probabilities.

  2. Gratifyingly methyl was tolerated says:

    And 1 pM against hERG

  3. Peter Kenny says:

    I loved “Wile E. Coyote physics” and I might start using it (with acknowledgement, or course) to describe thermodynamic analysis by the Budapest Enthalpomics Group (BEG). I don’t know what the bound conformation looks like but I’d guess that some of the magic in that methyl may be due to conformational biasing.

  4. Russ says:

    Derek, your rule about replacing nitrogens is not absolute. In the heyday of peptidomimetic chemistry my boss made a RGD antagonist with all the nitrogens replaced by carbon, and the compound was active at the receptor.

    Medchem, like organic chemistry the the science of exceptions to the rule.

    1. Derek Lowe says:

      That took some nerve! I did that once in a GPCR project, ripping out both nitrogens to produce what I referred to as my “nihilist anarchist compound”. But it was inactive, anyway.

      1. An Old Chemist says:

        I once replaced one Nitrogen with a Carbon, in a kinase inhibitor, and ended up having an inactive analog. The loss in activity was more thsn 20 fold. It was a multi step synthesis of the carbon analog!!!

        1. SedatedFMS says:

          I did something similar, got an active compound. A large number of analogues were made by me and others. There seemed to be a massive activity cliff when compared with the nitrogen containing analogues, the SAR just didn’t track. Comp guys examined the data, perfect correlation between the two individuals who ran the primary assay. Compounds ran by 1 generally active, compounds ran by 2, dead. The duplicate measures were run by the original operator. Retest ordered using a third scientist, the SAR was very weak. 25-30 compounds from multistep syntheses (ok some where from advanced building blocks) all because somebody couldn’t do simple dilutions correctly.

    2. MakeCarbonGreatAgain says:

      The cannabinoid agonist JWH-176 has all the usual heteroatoms replaced with C – still 26 nM.

  5. Anonymous says:

    This paper: Methyl Effects on Protein–Ligand Binding. Jorgensen, et al. J. Med. Chem., 2012, 55 (9), pp 4489–4500, DOI: 10.1021/jm3003697 says:

    “The effects of addition of a methyl group to a lead compound on biological activity are examined. A literature analysis of >2000 cases reveals that an activity boost of a factor of 10 or more is found with an 8% frequency, and a 100-fold boost is a 1 in 200 event. …” in the abstract, but further down is the key thing I remembered (my [emph] added):

    “The key observations from the survey are that on average in reported SAR series introduction of a methyl group is [[[ just as likely to hurt as help activity ]]] and that it is extremely rare for addition of a methyl group to give a free energy gain greater than 3 kcal/mol; in fact, only 4 of the 2145 cases are in this category. …”

    Of course, the data set will always be biased by what info the authors choose to disclose about their Me vs des-Me analogs. From published data, you can determine whether it is a Magic Methyl or a Miserable Methyl but you can’t say anything about the unpublished Secret Methyl.

  6. anon the II says:

    In all honesty, I don’t think I’d put this example in the “Magic Methyl” category. There are plenty of examples where adding a methyl kills the activity. It usually happens when you add a methyl and it bumps hard into a wall. The effect is what you’d expect from the left side of the Lennard-Jones potential. Like Anonymous suggests above, the “Magic” methyls are when you see the huge jumps in activity. They are much harder to explain in general terms, which is why they are “Magic”.
    Also, methyls are not really magic if they work their wonder by blocking metabolism.

  7. Matthew says:

    My understanding of a magic methyl is one that enforces a preferred conformation and results in a significant gain in potency. A nice example was described by Lodola et al (J. Med. Chem., 2017, 60, 4304) for PI3K delta inhibitors. Here the most potent inhibitor (the situation is complicated by a second methyl group producing atropisomers) is almost 100 times more active than the least. The parent compound without the methyl group is more than 10-fold down on the preferred isomer.

  8. Druid says:

    Are you med chemists given the precision of the assays? Unless you make a PAIN, the precision is a property of the assay which should be measured before you start inferring quantitative differences between analogues. In no other science would you consider differences without s.e.m.s to go with the numbers. Better than using an s.e.m. based on say 3 replicates, each assay has a positive control which gets used hundreds of times. In my experience, a factor of x3 or /3 is not unusual (95% CI), so a true 3nM could be 1nM or 10nM on another day. When you assay hundreds of compounds and several assays, there are sure to be some outliers which could be very misleading. At least put the number of replicates after the IC50.

  9. AC says:

    I can’t be the only one who first thought of the *other* kind of magic methyl known to chemists…

    1. tangent says:

      As a non-chemist who mostly knows the spectacularly horrible compounds, me too.

  10. Schinderhannes says:

    In my world Magic Methyl refers to Methyl fluorosulfonate! 🙂

  11. schinderhannes says:

    Oops AC beat I, next time I refresh before sending a silly comment after a break (in the lab).

  12. MFernflower says:

    Is it bad I thought of the rather nasty reagent with the same name?

    In my experience it’s always a magic CF3 never a magic methyl

    1. AC says:

      Or more generally a magic fluorine, really. Although that’s more for metabolic stability.

  13. Hap says:

    Another magic methyl: 10.1021/acs.chemrev.8b00422 (paywall)

    Ethylene carbonate (which melts at about 40C) forms a passivating layer on lithium-intercalated graphite, while propylene carbonate (melts at below -40C, and hence is preferred) does not. Since people assumed that the two would act similarly, it took a while for someone to look and find that they didn’t; this development was crucial to the development of lithium-ion batteries.

Comments are closed.