I’ve been thinking a lot about model systems recently, and it’s a broader topic than I’ve been giving it credit for. Back in graduate school, I used to think about model reactions quite a bit. I was doing total synthesis of an ugly natural product molecule, starting from things that you could buy in the catalogs. And by the time I got up to step 13 or so (not to mention step 27, which is about where I finally pulled the plug), the material I’d made was rather precious to me. And every time I started to do new chemistry on it, I always feared losing all my hard work. “Better a reaction that does nothing,” I’d think each time, “than one that chews up my stuff.”

So the idea of model reactions kept coming up. That’s where you don’t commit your bespoke, hand-made compounds to the new scheme right at first – instead, you make a simpler molecule with many of the same features (preferably, something very close to a commercial starting material), and try your new ideas out on it instead. This sounds appealing, but there are some pitfalls. The biggest of these is that the only real model for a complex molecule is that complex molecule itself. You can simulate some parts of its reactivity, but – as every experienced organic chemist knows – tiny changes can have very unpredictable consequences. That leads to a situation where you make the model compound, try your new reaction, and it works! So, you run your real molecule and hope for the best. Or you make the model compound, try your new reaction, and it doesn’t work . . . so you try your real molecule anyway, because you never know, right? An experiment that doesn’t change your course of action, no matter what result it gives, is almost always a waste of time. As is, most likely, the other possible way to proceed, which is to go back and make a new model compound, one that’s even closer to your real one this time, which will probably take longer to make, and can send you down a long, twisty corridor of diminishing returns.

Once I joined the drug industry, I stopped worrying about model chemistry reactions. For one thing, I wasn’t embarking on any 27-step syntheses, for which I am deeply grateful, which means that I didn’t have any compounds that felt as if my entire life were in solution with them. And I’ve rarely had any reactions exotic enough to require tippy-toeing up to them either, which is also fine by me. But the model systems I had to think about now were the assays in which my compounds were being tested. The main thing to worry about there was (and is) the animal models. That’s where the serious compounds have to go, when they have to be made in more serious amounts. And it can be a source of worry, once you start letting yourself think about it, to wonder just how closely some of these animal systems mimic the real world.

I started out doing a lot of work on Alzheimer’s, and perhaps that made me unusually sensitive to such things. After all, CNS animal models are notoriously wonky, and ours were no exception. We humans are still the only animal that gets Alzheimer’s, so it’s a real leap to go to any other species at all. And it’s hard to see how some of the models relate to Grandma forgetting her car keys, to be honest. When I look back on how many compounds I made, just to have them tested by seeing if mice ran into the dark half of an electrified cage when the light went on, it just makes me want to pound my head against the floor. (A chemist-pounding-the-floor assay, come to think of it, is not such a bad idea). The models for schizophrenia, depression, and many of the other big CNS indications are a bit better, but you wouldn’t want to bet the ranch on any of them.

In later years, I did oncology drug discovery, and found that I didn’t really trust most of the animal models there, either. Compounds that have gone to the market and basically failed (relative to their high expectations) can still perform wonderfully in xenograft tumor models. If only our customers were nude mice who were infected by tumors from a completely different species – we’d have had the whole field wrapped up a long time ago! But the real world of oncology is a much more complex and hostile place than the one of cell lines and compromised mice. Activity in one of those assays is in the “necessary but nowhere near sufficient” category, at least if you know what you’re doing. It has to be possible to make these systems better – you’d think – and a lot of effort has gone into trying to do so. But in the end, the only real model system for an aggressive cancer in a human patient is, well, an aggressive cancer in a human patient. Sometimes it has to be the exact same human patient!

But the latest thoughts I’ve had on model systems are even broader. Looking at the way that outsourcing of med-chem work has been going the last few years, it strikes me that, in some ways, I’m looking at small-scale models of regular drug discovery operations. The offshore suppliers will run reactions for you, crank out a series of analogs – but for many of them, that’s where it ends. Going on to test these in vitro – and especially in vivo – isn’t part of the deal. The larger outfits are moving into these areas, of course, since they’re higher up the value-added chain and more lucrative. And the companies that have their own internal offshore operations are also getting the whole package together. But still, in many of the cases, a company’s outsourcing operations are a deliberately simplified version of the real thing.

And these come with their own disadvantages, just as the lab-sized model systems do. The people involved may or may not have a clear picture of just where their compounds are going, or to what purpose. Some of the thought that could have gone into selection of analogs or prioritization may not get done, because of the compartmentalization I mentioned. There will, almost inevitably, be some wasted effort, all the way up to entire series of compounds that got made, but shouldn’t have been. The way to tighten these things up is, naturally, to make everyone more aware of all the details, but that’s often not a realistic option. And if you do choose to go that route, you might find that you’re making a new drug discovery organization from scratch, and perhaps not as low-cost an organization as you might have imagined.

So we’re back to the problem of making your model so much like the real thing that you might as well just have the real thing and be done with it. I wonder, sometimes, if that’s the way that we’re headed in this industry. We’ll look back, realize that we’ve taken a long, sometimes lucrative – and sometimes costly – trip, and look around to realize that we’re back in the same place from which we left. T. S. Eliot said the same thing, rather more poetically. Let’s just hope that the Eliot poem that best describes our industry doesn’t turn out to be The Waste Land, instead.

Derek B. Lowe has been employed since 1989 in pharmaceutical drug discovery in several therapeutic areas. His blog, In the Pipeline, is an awfully good read. He can be reached at [email protected].