Foreword to the 2nd Edition
Published:03 Feb 2023
I am very happy to have been given the opportunity to write this Foreword to the latest edition of The Handbook of Medicinal Chemistry. This science whose practice we all share has a long history and looks to have a lengthy future as well – but just as it does not resemble its beginning very much by this point, its later form will surely be quite different from what we see now. Large changes are coming. “Well of course they are,” you may well think, “there are always large changes coming”. What's so different now? What makes this Foreword different? Perhaps a bit of history will help make the case for that statement.
Targets and Drugs
My career as a medicinal chemist began in 1989, in a central nervous system research group. Naturally, I did not know enough to be terrified of the problems that were set before us, which were diseases such as schizophrenia and Alzheimer's. (It will immediately be noted, for example, that both of those are very much still with us and still in need of pharmacological aid). I soon learned, though, that my colleagues in the biology labs were quite excited about the G-protein coupled receptors that we had targeted in both areas, as new technology was displacing the tissue assays that had been in use for many years. It was, for example, farewell to the rabbit trachea and the rat vas deferens in muscarinic receptor characterization, and a look at the older literature1,2 will clear up any suspicions that I might be inventing those. Instead, we welcomed the cloned receptor subtype membrane-binding assays, which we felt would surely clear up many of our difficulties once we all got our bearings.
This switch to such individual targets with assays enabled by molecular biology had been taking place throughout the 1980s, with the late-1970s success of captopril3,4 as an angiotensin-converting enzyme inhibitor helping to light the way. But the idea of such targets themselves had been working its way into the practice of medicinal chemistry ever since Raymond Ahlquist proposed in 1948 that adrenergic signaling could be explained by what he termed “alpha” and “beta” receptors.5,6 The receptor concept caught on,7 but even up through the 1960s you could find warnings in the literature not to take such ideas too literally.8 That is, one was not to assume that these were necessarily defined molecular entities with real structures, but rather just useful mental models or a sort of shorthand to categorize the various mysterious classes of drug action. Even among those researchers who took receptors to have physical reality, it was unclear if they were proteins, complex carbohydrates, or who knew what. And as for how they might work? That was even more mysterious.
Receptors turned out to be real molecular species, of course – in fact, they were proteins that could be cloned and expressed once we knew how to do such things. The same went for enzymes, whose proteinaceous nature had become clear much earlier, but whose production and purification were likewise changed forever by molecular biology techniques. And thus medicinal chemistry went through one of its revolutions, with target-based drug discovery becoming the unquestioned default mode of the field. It still is. Target-agnostic phenotypic screening went into eclipse for a while, but even with its revivals over the years, the first thing that people will try to do with an interesting phenotypic hit is to … try to identify its specific target.9 And that's done partly so that the protein can then be expressed and turned into the subject of the non-phenotypic screening assay it would have been from the beginning, had anyone but known to look at it.
This is the world that readers of this new volume have been living in (as have I!) There's no doubt that it's been good to us. Whole classes of actionable targets have opened up over the years: enzymes such as the proteases, hydrolases, kinases, and phosphodiesterases, all ready to be inhibited, and also the receptors and ion channels to be addressed in all their profusion: agonists and antagonists, full and partial and inverse, all waiting to be discovered and exploited. A solid history of modern medicinal chemistry could be written from the perspective of watching these territories being mapped out and settled.
All of them attracted some doubters at first. With the profusion of kinase-targeting drugs that we have now, it can be difficult to remember that some very intelligent people wondered at one time whether a selective kinase inhibitor was even possible.10 After all, wouldn't you just hit every other ATP-binding site in the whole cell? Sometimes the doubters had a point, though: the nuclear receptors have had some striking successes as drug targets, but many of those came before we even understood them very well, and not so many have arrived since. Epigenetic enzymes went through a great period of enthusiasm too, but it's been a difficult field to make progress in. No doubt many readers can provide other examples from their own experience.
For a while, it was difficult to imagine that target-based drug discovery had any real limits, but anyone who took the time to think about it knew that it must. The number of pathways in living cells is large, but most certainly finite – as is the number of proteins, and of course the number of genes coding for them. Indeed, one of the most direct reminders that there would eventually be an end to it all was provided by the Human Genome Project. The speculative frenzy around human sequencing in that era was driven by the fear that unless you reached into your pocket immediately, other people (and not you, you cheapskate) would end up owning the rights to all of those targets and all of those proteins. Patents were being filed at a ridiculous pace, many of them doubtless overlapping in what promised to become an epic tangle during prosecution. The signs were clear: the great prairie of target space had at one time stretched almost unbroken to the horizon, but now there was dust flying everywhere as fence posts were pounded in and barbed wire was hurriedly strung up.
And then that horizon suddenly got closer. The initial 2001 figure of only about 30 000 human genes in total was alarmingly short of almost all the estimates at the time, and it caused widespread consternation.11 At that point, it seemed possible that more “genes” had been filed on than there were actual genes to be had, which is worth thinking about. At any rate, the whole patent-every-gene business model collapsed as it became clear that genes alone clearly weren't everything. Splicing variations and post-translational modification were responsible for much of the variety of the proteome, as we all should have remembered. It also became clear that the patent offices, in what has to be regarded as farsightedness, weren't going to even consider allowing the sort of patentability (and the sort of reach) for all those claims that people had been writing up.12
But bioinformatics was still showing everyone the end of the line. At one point, when gene sequencing was still so difficult and expensive, the question of how many human protein kinase enzymes there really were was basically unanswerable, at least if you wanted something more specific than “a lot.” But as the annotations and sequences got better and better, the answer took on an extremely discrete character: there are five hundred and eighteen of them. Similarly, the other target classes began lining up and counting off. It gave a person a strange sensation at the time.
There was still plenty of frontier, though, and for that matter there still is. It was obvious that we didn't know the actual functions of many of these proteins that had been enumerated. We still don't, not to the level of detail we would like. And there were big swaths of protein space whose annotations were as thin as tissue paper (and still are), with roles that have just barely begun to be understood. Eventually another shock will have to be faced when this unexplored intellectual space starts undeniably shrinking, but that time is not yet. It will come eventually, and though it's hard to say what the effect on medicinal chemistry will be, the field of particle physics over the last fifty years might offer a preview. At one of my former companies, we amused ourselves (in somewhat cynical fashion) by coming up with possible advertising slogans for our employer. One of my favorites was “(Name Redacted) Pharmaceuticals: Where Progress is Our Only Problem.” That joke, similar to a candy bar, had a chewy bit of truth in the center of it.
What have we done with all these target proteins? Generally, we have tried to stop them in their tracks. Let's be honest with ourselves: it's far, far easier to throw a wrench into the cellular machinery with inhibitors and antagonists than it is to make the gears turn faster by discovering enzyme activators or selective agonists. Most of the time, most drug discoverers have looked through the biochemical pathways of disease with an eye to which parts of them could be usefully disabled. Even current hot topics such as selective protein degradation (more on which shortly) are often just new variations on the “what can we interrupt” strategy. But all that wrench-throwing has been fantastically useful and informative, not to mention (on occasion) fantastically lucrative. A large part of the mapping and exploration mentioned above has been accomplished through the use of small molecule tool compounds, a favored few of which have become actual drugs. That process will certainly continue. A good small-molecule tool can shed an extraordinary amount of light on the biochemistry of health and disease, and can be made the subject of more and various sorts of useful experiments than almost anything else in medical research.13
It seems clear that there are currently more new modes of therapy dawning in the clinic simultaneously than there have been at any time in at least the past thirty or forty years. It's an amazing thing to watch. Not all of these are small-molecule driven (more on that shortly as well!) but some are, and others have a place for small-molecule targeting to assist them. We have CRISPR gene editing, chimeric receptor T cells (and other cellular modifications ex vivo), the above-mentioned targeted protein degradation, a revival of interest in antibody-drug conjugates, and likewise a growing interest in applying the techniques of organic chemistry to large biomolecules of many types.
That last field overlaps with the area now called “chemical biology,” and that brings up another point worth emphasizing. Medicinal chemistry (perhaps it would be more accurate to say medicinal chemists?) has sometimes had its scope limited by what can only be called a lack of imagination. We're not the only scientific field whose practitioners enjoy drawing boundary lines and declaring that the “real” subject only exists on this side of them, of course. But given med-chem's position at the intersection of several different sorts of biology and medicine, purity tests can be dangerous. Medicinal chemistry is where chemistry gets added to all of those other fields, and I've long maintained that if you have begun thinking of the subjects of your research as actual molecules, with shapes, sizes, charge distributions, distinct molecular weights, and all the rest of it, then you are turning into a chemist whether you realize it or not. That means that many molecular biologists are slowly becoming chemists, and I think we should let this process continue rather than scaring them all off for fear that they will dilute our own field. The era of ill-defined boxes, circles, and triangles with arrows pointing vaguely between them cannot end soon enough, and it will be finished off by understanding all these species at a molecular level. As chemists do. So my advice is for medicinal chemists to move into these new territories and bring their skills with them. We have things to offer that aren't available anywhere else, ways of thinking about molecular species that haven't yet taken hold in some of the other closely related fields and are very much needed.
Consider as an example targeted protein degradation, which as I write is attracting huge amounts of investment, interest, and labor. At the moment, the various chimeric species that are used to make TPD happen in living cells (or living patients) are somewhat mysterious. A great deal of experimentation is needed to come up with appropriately linked “business ends” to obtain the desired effects. We generally can't predict why some of these work so well (nor why others work so poorly), but it's clear that answers are out there for us if we can make our way to them. These will be found in the interaction of protein–protein surfaces, in the orientations of the target and ligase ligands, in the subtleties of entropy and enthalpy that go into the formation of a ternary complex and what we call cooperativity.
We also have a good deal to learn about the properties of the (often large and unwieldy) bifunctional molecules themselves. How do such species cross cell membranes, and what are their pharmacokinetic fates in general? We're not used to working with such things, and for now we are feeling our way. But such species, and some of the equally large and complex molecules that have made their way into the clinic as protein–protein interaction inhibitors and other classes, have (I believe it's fair to say) put a bookend on what has to be called the “Rule of Five” era.14 A renewed attention to molecular properties was probably a good idea, considering some of the excesses of the early combinatorial libraries, but most people who lived through the entire cycle would probably agree that the whole process went a bit too far.15 I believe that a department manager I worked with was joking when he wondered if there should perhaps be some sort of electric shock administered through the keyboard when a chemist tried to register some molecule with too many egregious rule violations. But another one was at least a little bit serious when he suggested a cap-and-trade system for the department, with every bench chemist getting a set quota of metal-catalyzed aryl–aryl coupling reactions per year.
Antibiotics, especially macrocyclic ones, were always known to be exceptions to most attempts to confine them to rules in property space. A great deal of effort has now gone into figuring out just why that might be so, with mixed results.16 But one undeniable result has been a broadening of medicinal chemists' thoughts, with large macrocyclic closures of all sorts making appearances in clinical candidates in recent years. Perhaps we all had to put on those tight buckles and straps for a while to see the best ways out of them. But the shackles do seem to have fallen off: molecules are moving into the clinic now whose structures might have gotten you fired twenty years ago – well, from a few of the more doctrinaire organizations, anyway.
This process looks to continue for a good long time to come. There are many other ways to apply the bifunctional-molecule idea, just to pick one large-molecule area. Instead of dragging over a ubiquitinating enzyme complex to a protein of interest, why not bring in any of the other post-translational enzyme systems? Why not use a bifunctional to nail down a particular molecule's location inside a cell, or to make some protein occupy one of those locations that otherwise it would never see? Why not move beyond proteins to glycosylated surfaces? Ten or fifteen minutes of staring out a convenient window will allow a person to sketch out enough research to keep everyone busy for years.
And this is just one area! A related possibility is in the intriguing area of “molecular glues,” small molecules that actually cause two otherwise unacquainted proteins to come together in unexpected ways. I will freely admit that such a mechanism would have seemed very weird and unlikely to me at one time, but here we are. For that matter, I would have furrowed my brow a bit at the bifunctional degraders, too, finding the whole concept a bit too unwieldy and brute-force. The poet Philip Larkin said that reading Thomas Hardy taught him “never to be afraid of the obvious,” and that's a lesson that we all get to learn, and relearn.17 Surprises wait for us at both ends of the scale, and the only way that we are going to run out of things to do would be through a failure of imagination.
The Shape of Things to Come
How we do those things, though, will surely be changing. Medicinal chemistry has seen any number of fads and periods of excitement over the years, as new techniques appear. From computational drug design (in all its forms), through combinatorial chemistry, fragment-based drug discovery, DNA-encoded libraries, and more, each of these has begun with a wave of enthusiasm (and skepticism) and has eventually found its place as part of the toolbox, another one of the many things that medicinal chemists do.
But we will certainly be making new molecules; that much is sure. The reactions we use to make them now are a mixed assortment,18 featuring some classic transformations from the earliest days of organic synthesis (such as amide formation) all the way up to the latest bond-forming catalysts from the literature. It has been demonstrated that a huge variety of diverse compounds can be prepared with simple chemistry using the great profusion of synthetic building blocks and intermediates.19 But at the same time, there is still a need for new reactions, especially those that might move into that same category of reliable bond-forming workhorses.20 Experienced medicinal chemists never lose sight of the fact that the chemistry itself is just a means to the end: producing a wide variety of new molecules for testing. If a new technique provides a reliable way to do that, and especially if it gives access to structures that were previously hard to synthesize, it will rapidly find a home in the labs. The relatively rapid uptake of photoredox techniques is an example, and there will surely be others.
I think, though, that we might be on the edge of some larger changes. And that takes us back to the assertions that I made in the opening paragraph of this Foreword. Let's consider what many of us have always thought of as the prerogative of a synthetic organic chemist: working out a synthetic plan to make a new molecule. Sometimes this is fairly trivial, sometimes very much not. But until now, coming up with a synthetic plan has involved one or more chemists sketching out structures and thinking about what can be made from what, and how. People bring their own experience to the process, of course, and they bring a good deal of literature searching for reactions and conditions. But in the end, it's always been a human process, full of human choices and human insights.
That's going to change. This is not a popular viewpoint, as I am reminded every time I express it in public. It's true that the various software packages that are now available to do retrosynthetic planning are often not very impressive. But what I find impressive is that they exist in the first place. And I feel certain that they will continue to improve, and to improve at a faster pace than humans do. Look, for example, at what has happened to games such as chess and Go, which were once only the province of human thought as well. There are no humans left that can beat the best programs for these games, and there never will be again. The same fate awaits every other game with set rules and strategy, and I maintain that organic synthesis is one of those games. Of course it's a game: that's why we love it so much. It has more pieces and more moves than chess, and it is played on a larger board than Go, but since when did such factors do more than slow down (for a while) a computational approach? No, for the great majority of synthetic planning, I foresee an era of what chess champion Garry Kasparov calls “chimeric” play, with humans and software cooperating to address the other's weak points.21 Eventually, I see the software doing more and more of the work and being able to fill in more and more chemical space with useful suggestions. It's admittedly hard to imagine a retrosynthesis program proposing a good route to palytoxin or vancomycin, but think about it: how often are any of us ever called to do such a thing ourselves?
As with the software, so with the hardware. People have been trying to automate organic synthesis for a long time now, but what some chemists might not realize is the extent to which that process has already succeeded. I refer to such familiar objects as rotary evaporators, fraction collectors, autosamplers, and the like. All of these labor-saving devices and more took over what used to be hand labor. Some of this has happened over the course of my own career: if you had shown me a modern chromatography setup, with prepacked columns running any sort of solvent mixtures from normal to reverse phase with any sort of automatically mixed gradient one might choose, with an option to only collect fractions when something actually seemed to be coming out the other end … well, I might have had to sit down for a while. It would have seemed an unimaginable futuristic luxury, but chemists in well-appointed labs now take this sort of thing for granted.
We may well be getting closer to more direct automation of the chemistry itself. We have increased use of high-throughput experimentation already, both in the analog-making and the reaction-finding senses of the term. And for years now, various groups and companies have been trying to automate the entire thing, with robotic weighing, mixing, reacting, monitoring, and purification steps. The dream has been to realize a fully closed loop, with some sort of automated primary assay joined to the synthetic process, and the results being then fed back, perhaps, to another piece of software which will recommend the next round of analogs for the synthetic hardware to produce.22 This may not to everyone's taste, but remember, dreams come in various types and with various degrees of enjoyment.
I'm not sure what I think about that one, either, at least in its strongest forms. But the advances in flow chemistry and in automating many regular laboratory tasks make this idea closer to realization than it has ever been. It remains to be seen how much sense (both scientific and economic) it will make for medicinal chemistry in general, but can anyone doubt that it might well be just the thing for the “go make me a hundred different amides” part of the work? That's not the most glorious feature of med-chem, true, but it's not to be sneered at, either.
And that brings up a more general point. Just as in other fields of work, all of this mechanical and electronic assistance is taking away what we would have called “grunt work”, while at the same time gradually expanding our beliefs about just what constitutes grunt work in the first place. Synthetic chemists used to have to blow all their own glassware and synthesize all their reagents, little of either being commercially available. I thought nothing about packing my own chromatography columns and collecting all my fractions by hand when I was in graduate school, but many chemists now would feel differently and I'm with them. There are better uses of one's time. As the machinery advances, we will all in fact have to find those better uses. We will be gradually moved, perforce, into the areas where the robots and the software packages cannot yet help us, into higher-order problems that are less open (for now?) to computation and automation. Synthetic organic and medicinal chemists will need to get used to working at a higher level, since the lower levels will, as the years go on, be less and less our concern.
If that really is the way the future goes, then the field will end up eventually in a very different shape than it is now. When I started in industry, there were older chemists around who'd been knocking around the labs for decades, stretching back to an era I had trouble picturing at all. Some of them were content just to keep their heads down and crank out molecules at the bench – I well recall one of them telling me that he didn't want the spotlight and didn't want to be working on anything flashy or high risk. He just wanted to come in and do his work quietly and reliably on whatever ordinary chemistry needed doing. For better or worse, this type of chemist has been harder to find in medicinal chemistry departments over the years, and they look to get even rarer still. I don't foresee everyone getting more specialized, though. The jack-of-all-trades, the débrouillard, that sort of chemist will always have a place and may well be needed more than ever. But the trades such a person will divide his or her time among and the problems where ingenuity and insight will need to be applied – those will be different. They already are.
To borrow a line, it's not going to be chemists versus machines. It's going to be the chemists that use the machines productively versus the ones who don't. We're going to need all the help we can get. As we move into the sorts of complex mechanisms discussed earlier, it looks very much like we may be making larger, more complicated molecules, and plenty of them. We may well find ourselves spending most of our time on more complex chemistry than we tend to now, and the idea of setting up a set of amide formations yourself in a fume hood might start to sound as antiquated as going out to collect firewood.
Readers will have noticed that I have not until now been talking about molecular modeling and computational drug design per se. The number of failed predictions and retrospectively ironic headlines in this area gives me pause, and I don't want to add any more to the list. No other part of drug discovery seems to have had as many repeated bursts of enthusiasm, and while that has made for a bumpy ride, it remains true that there has always been a lot to be enthusiastic about. I'm fond of saying that I'm a short-term pessimist and a long-term optimist about computational drug discovery – by that, I mean that I see no reason why it eventually won't be extremely useful, but if someone walks up to me next week and says that they have solved its problems and everything is ready to go, I will still hold on to my wallet.
The timetable for when the switchover to “extremely useful” comes has slipped several times over the years, but there's still no doubt that modeling, docking, virtual screening, and all the other techniques have made a lot of progress. And it's certainly not the modelers' fault that progress has been harder to realize than people thought it would be back in the 1980s. But while (say) calculating small-molecule ligand binding to anything close to experimental error is a hard problem, it certainly doesn't seem to be an insoluble one. Betting against a field that has the increase of computational power on its side and can be advanced at any time by sudden algorithmic insights is not a wise move.
There's also the advent of machine learning and AI to consider. Even as I write this Foreword, word has come of the best results yet on calculating protein folding, and it came via an AI approach.23 That result illustrates the strong and the weak points of such systems. Thus far, they are very good (sometimes terrifyingly so) at sorting through existing knowledge. The protein-folding result, for example, was not realized so much by fundamental insights into hydrogen bonding or torsional strain, but through advances in recognizing protein structural motifs in the huge archives of existing data. The application of such techniques to the retrosynthesis programs mentioned earlier should be obvious (and this is one of the reasons I think that they will be increasingly useful). What an AI will not do much of, not yet, is to help you out with things that are not already buried in what's known. But that still leaves a great deal of scope for interesting things, particularly as AI algorithms go to work on the problem of optimizing themselves.
The End Results
But never forget, all this is in the service of great things. We medicinal chemists work on so many failed projects of various kinds that it can be difficult to stand back and look at some of the successes. Oncology is a dramatic example. Thanks to new approaches in that field, there are people out there today deciding what to order for lunch who would not even be among the living under earlier treatment regimens. The startling inside-of-a-year advent of vaccines against the SARS-Cov2 virus is another case in point. Vaccine development will clearly never be the same again, and the world will be the better for it.
The central fact of drug discovery, and of the drug industry itself, has been the horrendous failure rate in clinical trials. For many years now, it's stayed around 90% over the whole field, with some therapeutic areas doing better (cardiovascular, infectious disease) and some doing worse (oncology, CNS).24 But think about it: if we could only manage to fall on our faces eight times out of ten, we would roughly double the number of new drugs making it to patients. That would be a tremendous outcome, and it isn't going to take anything miraculous to achieve it. A steady pace of improvements up and down the process, even though many of them might be small, will be enough. We can heal the sick, we can improve people's lives, and once in a while (as with some of the recent advances in immuno-oncology) we can nearly pull people out of the grave. Every year we get better at it; every year we know more. It is, when you think about it, a great job to be doing.