Skip to main content


Replacing Antibodies With Small Molecules

As anyone who’s been following the oncology field knows, antibodies against either the PD-1 receptor or its ligand PD-L1 are about the biggest things going in the field right now. Hundreds of clinical trials are underway against various tumor types and in various combinations, in the effort to see how far the immuno-oncology idea can reach. But do the drugs have to be antibodies? Is there room for small-molecule compounds in this space?

That’s a big question, given the number of antibodies that are on the list of best-selling drugs. People tend to assume (both on first principles and based on prior experience) that the sorts of protein-protein targets that antibodies tend to fill are next-to-impossible for small molecules to deal with, but that’s not necessarily true. But how often it isn’t true, how you can identify those cases, and how much time and money it would take to work on them. . .those, no one is quite sure about. The advantages of such compounds would be that they would have (presumably) more “typical” pharmacokinetics than antibodies do. The latter tend to linger – the PD-1 antibodies are given once every few weeks, for example. That means that if there’s a bad reaction to an antibody, the patient is just going to be experiencing it for however long the drug takes to clear out, as opposed to the once-a-day dosing typically seen with oral small molecules. Dosing is also less complicated as well, without the need for i.v. infusion.

What’s for sure is that people are trying small molecules in the PD-1 space, as shown by this review. (Interestingly, you have a choice of several other reviews of the exact same topic in several other less-prominent journals). There is an X-ray crystal structure of the PD-1/PD-L1 binding pair, which is a good (probably an essential) start to such efforts, and it’s one of those good news/bad news situations. The good news is that there are some polar-looking hot spots in the binding surfaces, but the bad part is that there’s a lot of hydrophobic surface binding as well (plenty of isoleucines pitching in, not an uncommon situation, and you don’t get much greasier than isoleucine side chains). For some recent reviews of targeting protein-protein interactions in general, see here, here, and here.

The obvious first thing to try in such situations is a peptide-ish compound that can mimic as much of the binding partner as you think you can get away with, but that’s a lot easier to write than it is to do. If you just administer the minimal peptide itself, it’s going to get destroyed, pretty much every time. “De-peptidizing” is definitely possible, but it’s pretty empirical, and there are a lot of choices, so you’d better be prepared for a long haul. The review goes into several attempts in this line. One of the most popular strategies is to go after macrocyclic peptides, which can have better stability and better absorption and cell penetration. Or not! That takes you right back to the “pretty empirical” space, because despite a great deal of effort, there still seem to be no general rules about what sort of macrocycle you’d want out of the vast number of possibilities.

You can also just settle in for a small-molecule screening effort, but such campaigns for protein-protein interactions tend to have low hit rates (and plenty of false positives). But like depeptidization, it can be done, and has been. Some of the most well-studied compounds in this space are a series from Bristol Myers-Squibb (as published on by researchers in Finland, Poland, and the Netherlands). These turn out to bind to the PD-L1 protein and apparently induce dimer formation, making it unavailable for the PD-1 receptor. Aurigene and Curis are also working on some compounds (CA-170 and CA-327), whose structures have not been officially disclosed, but which appear from patent filings to be broadly in the peptidomimetic class and are in early clinical testing. If there are other small-molecule efforts in this area (Merck?), they have not surfaced publicly, as far as I know.

The general questions mentioned above are still very much open: how often can an antibody be mimicked by a small molecule drug? And how effectively? Given the huge variety of antibody binding motifs and protein-protein binding in general, there probably isn’t a general answer, but an endless series of “Well, it depends. . .” No one’s happy with that sort of answer, because it can also just be a way of avoiding a question (or avoiding the work needed to answer it), but at the same time, in this business that’s often the only real answer that can exist. It’s unfortunate that delivering the truth can sometimes make you hard to distinguish from someone who’s trying to avoid doing so. But for now, a big consideration for someone wanting to do a small-molecule antibody replacement project needs to be “How much do you want one”? Because even if it’s possible, it’s not going to be fast or cheap, so you’d better be sure that you really want the tiger that you’re going to spend all this time and money stalking after.

Addendum: although the review article linked to above seems to be a good overview of the topic, to be honest it could have been edited more tightly for grammar. Too many articles are dropped, making the sentences rather choppy to read, and there are some odd word choices. For example, “Compared to CTLA-4, inhibition of PD-1 pathway has chronically stimulated state; therefore, anti-PD-1 drugs own increased antitumor activity and relatively favorable toxicity profile” and “The aberrant expression of PD-L1 and PD-L2 in tumor cells impairs body’s antitumor immunity“. The sections describing the various molecules reported are written more smoothly than the introduction, though.

18 comments on “Replacing Antibodies With Small Molecules”

  1. luysii says:

    Interesting to see how Derek’s recent posts have been more about the biologic systems drug chemists are trying to change using drugs than pure chemistry itself (Ugi reaction 27 September). This is not a criticism, and you can’t understand (or find) the drugs without chemistry, but you have to know how what you’re trying to change works.

    While we all love organic chemistry, physical chemistry, polymer chemistry etc for their own intrinsic beauty, they are also means to an end. This keeps them focused to some extent. For what happens when a field is driven by its own internal dynamics, read Woit’s blog ( or Hassenfelder’s blog ( about string theory and physics

  2. MrXYZ says:

    On the antibody side, companies are getting better a making antibodies against proteins that are traditionally drugged with small molecules (such as GPCRs). So the question is when to use an antibody versus when to use a small molecule. As small molecules and antibodies begin to encroach on each other’s territories, I suspect that it really will come down to dosing and PK issues (as discussed in the blog post). For antibodies, the lack of BBB penetrance will give an advantage, in some cases, over small molecules for certain GPCR targets (where you want to avoid CNS effects). The ability to dose less frequently with antibodies is probably a double-edged sword: useful for chronic conditions but less useful if tox issues develop and you need to get the antibody out of your system quickly.

    Any good reviews on how to think about choice of therapeutic modality?

  3. drsnowboard says:

    Any comment on Warp Drive Bio being shuttered / reversed into Revolution?

  4. Chemist Turned Banker says:

    I wonder if another consideration is project value? If you produce a mAb for PD-1/PD-L1, you have 12 year’s exclusivity and, as it stands, non-interchangeability of biosimilar that might be at a ~30% discount, so the post-patent tail of the molecule is not insignificant. If it is a small molecule, it will be savaged as soon as the patent has expired. The difference in NPV could be quite dramatic.

    As an organic chemist by training, this gives me no pleasure, but I’m sure it’s a factor

    1. chimaera says:

      On the flip side, production costs for small molecules are much lower, and dropping as continuous processing technologies are being brought to bear on pharmaceutical production. Production rate and therefore volume are also larger.

      High volume + low cost of production = lower cost price.

      Lowering the cost price increases the market opportunity a lot, particularly if you can target a disease that’s not worth pursuing with expensive antibody based treatments.

  5. Thomas Lumley says:

    The other advantage or disadvantage of small molecules, depending on your point of view, is that they eventually go generic and become less profitable.

    1. Thomas Lumley says:

      Ah. Got beaten to that point.

  6. An Old Chemist says:

    A few antibodies do cross the BBB. For example, lately a lot of monoclonal antibody drugs were in news after they failed clinical trials for Alzheimer’s disease.

    1. Rigorous science for all says:

      … and what was the evidence of crossing the BBB? CSF concentrations of the antibody?
      If that alone, I am not convinced they got in the brain reached the target…

      1. MrXYZ says:


        If you look at papers on antibody PK, roughly 0.1% to 1% of antibody gets into the CSF. The actual amount that gets into the brain has to be way less than that. The only exception might be cases where either the BBB is disrupted by disease or the antibody is specifically targeted to cross the BBB (e.g. using transferrin receptor as a shuttle – demonstrated in animals but not sure ever put into the clinic).

        I am not sure any good evidence for crossing the BBB in any of the Alzheimers papers though I could be wrong on that.

        1. Lane Simonian says:

          Damage to the blood-brain barrier is indeed likely the way that a number of these antibodies reach the brain.

          1. Rigorous science for all says:

            If so, antibody brain concentrations in models of compromised BBB should be higher than in the non compromised BBB controls. Are you aware of any such studies?
            And if so, what happens to the CSF concentrations upon BBB disruption?
            Thank you.

  7. Ingo Hartung says:

    Next generation immunotherapy treatment regimes will likely comprise combinations of drugs targeting different stages of the immunity cycle, e.g. aPD-1 plus an agent boosting antigen presentation/T cell priming. In such combination regimes SMOLs may turn out to be especially advantageous due to their higher dosing flexibility. Either to stage one drug after the other or to titrate a SMOL on top of aPD-1 with the goal to minimize irAE. Keep in mind that the combination benefit of aPD-1 plus aCTLA4 is limited by a significant rate of high grade AE and many patients discontinuing treatment. The other obvious benefit of SMOLs would the ability to address signaling nodes downstream of the T cell receptor or inhibitory co-receptors.

  8. hn says:

    A few years ago, I took a quick stab at virtual screening for compounds that would disrupt PD-1/PD-L1. Hits scored poorly presumably because of the shape of the interface. Most were peptide-like. None of the ~10 hits I tested showed activity in an assay.

  9. Bertrand says:

    How about aptamers (or better, somamers as developed by SomaLogic)?

  10. Lane Simonian says:

    Ferulic acid may be one of the best small molecule candidates for the treatment of cancer. Ferulic acid not only inhibits a number of the pathways leading to tumor formation, it also limits tumor immune escape.

    1. mfernflower says:

      I’m worried about FA being a possible PAINS

  11. mfernflower says:

    Why not simply make a antisense oligonucleotide of whatever codes for PDL1?

Comments are closed.