Skip to main content

Analytical Chemistry

Knowing the Structure

There are a lot of topics that we really should know more about in drug discovery, but which are buried inside projects inside particular organizations. Some of this knowledge is available once you’re inside said organization, but some of it is hard to assemble even then (much less into a review from outside). An example is the impact of structure-based drug design. There’s no doubt that it has had a major influence in some projects, but it’s hard to quantify.

There are, of course, projects that can be largely ruled out, since there isn’t really a structure for the protein target (and thus certainly not one with a ligand bound). But that still leaves you with a lot of projects where effort goes into getting such data: if you have a protein that can crystallize, it’s really negligent if you don’t give it a shot. But I’m not aware of a recent review or list of structure-based successes (where key insights were gained this way), even though I know that they’re out there, and have been involved with some myself. These tend to be a paragraph inside a longer paper, when the project gets published at all, so it’s no easy job to track them down.

With that in mind, I’d like to throw this question open to the readership: what are some recent examples? My guess is that fragment-based drug design is where many of them will turn up, since you really want to know the details of your fragment binding modes. But there must be many times where a particular methyl group’s or hydrogen bond’s role became apparent in an X-ray structure and informed the rest of the SAR development in a crucial way. I’d be interested in hearing if anyone has been able to get this sort of thing done by NMR techniques as well. In a few years, we’ll be asking about the influence of ligand-bound structures via cryo-EM, the way things are going, but for now I’ll bet it’s almost entirely X-ray work.

So what’s the state of the field? Is the value of structure-based methods so apparent that no one notes it so much anymore? Are there programs that got ligand-bound structures but found that these provided no particular insights? Or are there some clear success stories that have been highlighted but not collected into any one place? Examples welcome.


20 comments on “Knowing the Structure”

  1. Peter Kenny says:

    I would guess that the availability of structural data has led to an increase with time of the molecular size and complexity of clinical candidates. Like Derek, I would also guess that fragment-based drug discovery would never have taken off without structural data since medicinal chemists would not have had the confidence to pursue low-affinity starting points without this information. These days NMR appears to used for hit detection rather than for structure determination.

    Increased availability of structural data has led to partial eclipse of some computational approaches. For example, building a pharmacophore to infer bound conformation and binding mode is a pointless exercise if you have a structure of the protein-ligand complex. Docking-n-scoring can be applied in the absence of a pharmacophore model.

  2. Fragment-based drug design is often thought to require a structure, but in a poll last year (link in name) a quarter of respondents would be willing to begin optimizing a fragment without structural information.

    That said, the majority of FBLD programs described in this recent review were structurally-enabled – details in the paper.

    J. Med. Chem., 2018, 61 (5), pp 1774–1784
    DOI: 10.1021/acs.jmedchem.7b01298

    1. Peter Kenny says:

      Hi Dan, I don’t think there’s a simple yes/no answer to the question of whether or not one would start an FBDD project without structural information and a lot depends on other factors. The combination of relatively potent fragment hits (say 50 µM) and an assay that can reliably measure weak affinity enables you to develop SAR and gives you room to maneuver. If you have a single 500 µM hit and an assay that just tells you that it binds then life becomes a lot more difficult. The value of the target is also relevant and I recall a number of ‘must win’ projects from my time in Pharma (strangely enough we never seemed to win but this never seemed to trigger the culls in the managerial ranks that one would have expected from not winning when winning was compulsory)

  3. Curious Wavefunction says:

    Lots of recent examples from the kinase and HCV protease and polymerase fields (including FBDD), and I would like to see a review myself. Anyone willing to join forces on writing one?

  4. Chrispy says:

    My experience at a large pharma which enthusiastically pursued structures wherever possible (and even in some cases where it was really not useful, like GPCRs) was that by the time the structure was solved the chemists has already worked out the SAR empirically. The structures did make for some nice PowerPoint slides, though!

    Interestingly, structural information did come in useful for antibodies, not for making them better but rather for defining and patenting the bound epitopes.

    1. bhip says:

      GPCR crystal structures hasn’t helped?

      1. Dr CNS says:

        Not really. To my best knowledge, functional in vitro assays are the real breakthrough for GPCRs.
        Of course, some companies do use it (Heptares) and learn from it interesting lessons.

      2. Chrispy says:

        No. GPCRs are innately really floppy, and the crystal structures are not very informative. Plus as a class they are generally very druggable, so as Dr CNS suggests, it is really the screening that is the most helpful.

        It probably didn’t help, too, that by the time a GPCR crystal structure was solved you inevitably had a bunch of detergents and you were probably using a sequence from some non-human species, anyway.

    2. Chris says:

      In the field of antibodies, structure has played a critical role in the design of byspecific antibodies. The idea behind Genetech Knobs and Holes design came from examining the heavy chain interface of an antibody’s Fc domain. Many of the other bispeicifics designed such as Charge inversion, SEED and others have all started from examing the Fc domain crystal structure.

    3. Barry says:

      If Structural biology is to be useful in the Med. Chem. (rather than just a pretty image when the project is over and published) protein expression/purification for crystallography has to begin at the start of the project, before hit-to-lead. Even then, of course, no one can tell a novel protein (or a chosen domain of a protein) when to crystalize.

  5. Dominic Ryan says:

    This question has been asked regularly for the last 20 to 30 years. The answers have consistently been inconsistent.

    It is true that in big pharma, historically at least, there have been projects where the emergence of a structure did not actually add significant new information. That is not a failure of the utility of structural biology. That is a failure of project management for not having kicked off the structure work early enough or having understood the opportunity for the structure class!

    Nor should one divide the information into what you get from a structure vs. what you get from computational chemistry (which includes but is not limited to docking and scoring). The protein-ligand crystal structure is one useful snapshot but not the whole picture. I had one project where the key ligand interaction was ‘fuzzy’, it turned out to be due to needing a better construct with a partial loop deletion and improved unit cell packing. Modeling and SAR helped to drive that clarification but structure was needed to nail down some key uncertainty.

    At Cubist we ran lots of fragment screens by NMR and they were very useful indeed. NMR fragment screening complements crystallography. Results from one often do not reproduce in the other. That is not because either one is wrong though. They are different pictures.

    I think that the question Derek is asking is not actually about structure. It is about project teams. The comparison really needs to be a four-way comparison. Two ‘good teams’ that do or do not have structure and two ‘poor teams’ that do or do not have structure. Of course all on the same project.

    The ‘good team’ with structure will make good use of it and likely make better decisions than without. The ‘poor team’ might get rescued sometimes but might also get misled at other times. Now to answer the question all we need is enough examples to achieve statistical significance. Think of it as a clinical trial.

    The much more important question is how often a clever 2D pulse sequence has produced a clinical candidate? 🙂

    1. Former NMR Person says:

      I would venture to suggest the answer to your question is most of the ones that reached that stage in the last ten years. I doubt many molecules make it to the clinic these days without some use of 2D NMR somewhere in the project. The sequences used may now be considered routine but they were definitely cleaver when they first appeared. In my view the fact that they are still being used is proof of that.

  6. not sure says:

    I am crystallographer in pharma. So my view could be biased. Structural biology is a powerful tool for med chem, but only if you use it right. It typical takes time to set up, and it requires close collaboration between structural biologists and chemists to maximize its potential. A structural biologist should fully understand what the chemistry team is trying to do and what can be learned in structures that would be valuable for the team. Unfortunately, not every structural biologist is capable of doing that or even cares. And on the flip side, many chemists treat their structural biologists as in house CRO and are unwilling to get them involved. So its not surprising often times it didn’t work as we wished.

  7. John Wayne says:

    Derek, it may be impossible to really answer your question. The behavior of research groups are constrained by resource availability and what tools are considered optimal by your organization. This probably dominates any sort of academic questions like, ‘if you could do what you wanted, when would you use structure-based approaches to medchem?’

  8. milkshake says:

    here is a bizarre example how it works: Crizotinib (Xalcori) is a billion USD cancer drug. It started as oxindole Sutent-like lead for c-Met kinase at SUGEN. The PK property of the lead were terrible (injection site precipitation, unpredictable metabolism). A new aminopyridine lead came from HTS at Kalamazoo after Pharmacia bought SUGEN. The company eventually co-crystallized both leads in the active site of c-Met, and combining the structural info from oxindole series helped to dramatically improve the aminopyridine series. This time the lead was fairly potent and had good PK properties but it had poor target selectivity. This was considered a plus, a competing second series with exquisite c-Met selectivity was eventually shelved by Pfizer after Pharmacia merger. The original research team was mostly laid off by Pfizer and transferred to San Diego (former Agouron), only the project manager stayed. There the project languished and was eventually shelved amid the post-merger chaos. Later on, a biologist at Pfizer was trying to find medchem support for ALK, and since the binding sites of c-Met and ALK are somewhat similar, he had the c-Met series re-screened, and bingo – crizotinib – which was taken into clinic without any further modification since the animal PK data a potency in ALK was good (poor selectivity and activity in c-Met was again considered a plus).

    So we have a perfectly rational old-fashioned structure-based drug design, that was done in a perfectly rational way – for the wrong kinase – and was repurposed by a smart biologist taking advantage of the off-target activity…

    Oh, and that c-Met project manager at Pfizer, she took all the credit for herself, republished the research of her SUGEN colleagues as her own, without putting their names on the publication. She was enormously promoted, became Pfizer Fellow and then went on to start her own biotech company.

    1. Curious Wavefunction says:

      Interesting since crizitinib was in fact developed initially for c-Met as the target. Hundreds of cell lines with c-Met mutations were screened; one of the cell lines was NSCLC with ALK mutations. The drug was then repurposed to target ALK.

    2. hn says:

      Fascinating story. Thanks for sharing!

    3. anon says:

      Last time I checked, he did not change the gender

  9. Omar Stradella says:

    We were working in a collaboration with an academic group. They had a series of inhibitors of an enzyme and we had another series from our own screening. It was impossible to compare the SAR from one series with the other until they were able to obtain co-crystals and solve the structure. It turned out that 3 floppy phenylalanine rings facing the inside of the binding site changed conformations so much with the different inhibitors that they almost completely reshaped the binding site. After that, some limited structural cross-pollination was now possible.

  10. lxc says:

    This is something that entry level crystallographers like myself 2 years ago find very frustrating.
    There are no or few papers that I know of that describe the process of structure-based design accurately with all its frustrations, confusions, rabbit holes, surprises… Most present incremental positive changes in potency and phys-chem properties, which hasn’t been the case in real life. Only a handful of the compounds that were made and tested are presented (out of hundreds). This is also the case in conference presentations. Every senior crystallographer I talked to about textbooks or the literature has told me not to bother looking because most of the learning will be done on the job; except for a few papers on SBDD and FBDD with mostly Kinases as targets.
    At my company, work on Xtal structures starts as soon as a target is nominated as a potential project. For a couple of my projects, the structures were part of the hit validation funnel before contributing to lead generation and optimization.

Comments are closed.