Robert Plenge has an excellent post here, drawing on this recent paper from authors at Stanford. It’s on the idea of polygenic traits and disease, a very worthwhile subject considering what’s going on in the drug industry these days.
I say that because I’ve been making the joke, for some time now, that if you were to step out into the street in Cambridge and yell “Friedreich ataxia!” that about a dozen people would probably turn around, because they’re working on it themselves. That goes for several other diseases, and why they have in common is that they’re more or less localized to particular genes. In the case of Friedreich’s, it’s the gene for a protein called frataxin, and for Huntington’s it’s the gene for huntingtin, for cystic fibrosis it’s the gene for the CFTR protein, and so on and so on. These monogenic disorders have a reasonably clear biological picture, up to a point: you at least know exactly what the target is, and if you can fix the problem with that target protein, you are almost certainly going to have an impact on the disease.
And that’s good, because these diseases are severe and basically untreatable in most cases. The first disease-modifying therapy that comes along is going to have a very big impact, and stands to have a very good chance of FDA approval, if the side effects aren’t too severe. I should also note the rare-disease pricing model, since it’s clear that insurance payers have been willing to tolerate high prices for such rare diseases (and indeed, it would be financially impossible for a company to discover and develop such therapies for relatively few people if the prices weren’t high).
So why hasn’t everyone been stepping right up for all this easy money? There are two answers – they have, and they haven’t – and they’re both true. Historically, these diseases haven’t had as much work done on them, because of that small patient population problem (among other things – see below!) But with the advent of the current pricing regimes over the last ten to twenty years (pioneered by Genzyme, among others), that’s changed. The last few years, it seems like everyone has been piling into this financial model, thus the shout-on-the-street line.
The catch is, though, that these diseases and their protein targets are not simple things to work on. The long-term solution, clearly, is modification of the human germ line, through CRISPR or some other DNA-editing technique – then no one will pass on any such defect to the next generation. But we don’t know how to do that safely enough and specifically enough quite yet, and although I think such a procedure is going to eventually come along (and be accepted), a lot of people are understandably jumpy about permanent changes to the human gene pool. The second best solution is to modify the relevant cells in a patient who’s already been born with the condition. That’s being tried as we speak in sickle-cell anemia, because the stem-cell population that produces the aberrant red blood cells is well defined and can be replaced through bone marrow transplantation.
As an aside, here’s an overview of the state of sickle-cell anemia work from Sharon Begley at Stat. I think I’m less upset and shocked than she is about the rate of progress in the field, though. From my perspective, every single time a new technique comes along that might be applicable to a genetic disease, sickle cell has been the proving ground for it, because it’s so well defined. The relative lack of progress against it, I think, reflects how difficult the problem is. It’s true that a “moonshot” blast of funding (how I hate that word) would likely have sped things up, but on the other hand, there are only so many of those to go around, and they come with no guarantees attached. Attacking the defective hemoglobin gene or the defective proteins themselves, or getting the useful fetal hemoglobin gene to turn on instead, are not problems with any sort of obvious solution. In recent years, another possible target in the pathway (Bcl-11A) has come up, but that’s no slam-dunk, either, since you’d need a good, selective protein-protein inhibitor. (That, by the way, is often the case with these diseases – even if you find another way into the problem, it’s usually a protein-protein or transcriptional regulation pathway, both of which are going to take plenty of brains, work, time, and money to attack).
Remember, sickle-cell is considered the straightforward one. Other genetic diseases can be harder to deal with. Take Friedreich ataxia, for example. It’s clearly linked to defects in the frataxin protein, as mentioned, but frataxin is found all over the body. You’re not going to just be able to step in and transplant the problem away, not when it’s in the brain, liver, kidney and other organs simultaneously. So now you have to work at the level of the protein target and find some way to get gain-of-function (or at least mitigation-of-defect), which is not easy. One big complication is that we actually don’t understand many of these proteins even in their fully functional state – what, for example, does frataxin do, exactly? Well, it’s pretty clearly involved in oxidative stress in mitochondria, so far, so good. But how? Something to do with iron? Superoxide radicals? Partnering with other mitochondrial proteins? You get to the hand-waving stage very quickly, unfortunately. How do you fix something when you don’t understand what you’re trying to fix? People are working on just that problem, as well as trying to find side routes and bounce-shot ways into affecting the disease process, but don’t expect anything real soon.
This is the big catch in the rare-disease gold rush: the valuable nuggets are fiendishly difficult to extract, in most cases. What everyone would like is a serious disease, thus having a high unmet medical need, with a clearly defined genetic target and mechanism of action, and one that can be addressed through the traditional therapeutic avenues (inhibition of some enzyme, blockade of some membrane protein). But those are mighty thin on the ground. Everyone wants to buy some stock that will rise quickly and reliably with little downside risk, but those are kind of hard to find, too.
And that finally gets me to Plenge’s blog post and the underlying paper. By its classification, I have been describing the easy parts of the genetic approach to human disease. These are what the paper calls “core genes”, ones that are clearly related to disease mechanisms and can provide insight into the whole etiology. But there simply aren’t that many monogenic diseases out there. Beyond those, you have “peripheral genes” that are particularly important in certain cell types, and for which we understand broad categories of function, but which don’t rise to the “smoking gun” category of (say) the defects in hemoglobin for sickle-cell disease. And past those, there are vast, huge numbers of genes and pathways with nonzero contributions to disease, but which pile up and interact with each other in ways that we truly are only barely beginning to sort out. As the paper says:
In summary, many complex traits are driven by enormously large numbers of variants of small effects, potentially implicating most regulatory variants that are active in disease-relevant tissues. To explain these observations, we propose that disease risk is largely driven by genes with no direct relevance to disease and is propagated through regulatory networks to a much smaller number of core genes with direct effects. If this model is correct, then it implies that detailed mapping of cell-specific regulatory networks will be an essential task for fully understanding human disease biology.
I’m afraid they’re right. This is why the searches for the “Alzheimer’s gene” or the “schizophrenia gene” were futile, because things just aren’t that simple. Genome-wide association studies are not worthless, far from it, although there sure are some worthless ones out there in the literature. But there are clues that they can give you and things that they’re going to leave blank. One of the key factors, as Plenge’s post makes clear, is that magnitude of the effect from a given biological change. This can be tricky to work out; it’s a constant question even in non-genetic-driven drug discovery. What’s the “tone” of the system – how much do we have to inhibit the function of Enzyme X or block Receptor Y to have a real effect on the disease? In terms of most rare-disease projects, how much of the function of defective Protein A do we have to restore (by one means or another) before the patient gets better?
Thus, it is essential to think quantitatively about the biological effect of a disease-associated variant, not just the fraction of heritability explained by the disease-associated variant. This distinction is what seems to confuse a lot of people. As I recently attempted to do for the TNFSF13B gene / BAFF protein (see blog here), drug hunters need to know the magnitude of effect of perturbing a target to decide if it is a good drug target or not.
As an aside, a pet peeve of mine is when someone comments: “I don’t care about variants with odds ratios of 1.10.” This is a nonsensical statement, unless you know the biological effect of the disease-associated variant. As an extreme example: consider a variant that completely obliterates gene function and has small clinical effect. This observation would effectively rule-out a target for pharmacological perturbation. Oh naïve drug hunter, you should care about this finding!
This is absolutely true, although not always a popular statement to be making when there are projects to launch, targets to hit, goals to be made. You’re basically saying “No, no, wait, we have to look at this more”, and people get suspicious (for good reasons) when that always seems to be the answer. But that’s the hard part: sometimes it really is the best answer, just as sometimes plunging ahead and damn the torpedoes is the best answer, too. Those calls are what people get paid the sorta-big bucks to make.
People – managers, investors, reporters, doctors, the general public – get tired of hearing “Gosh, it’s more complicated than it looks, we have to do more work”. But all of this stuff really is more complicated than it looks, and when it already looks complicated to start with, well. . .Here’s the Stanford paper again:
Huge numbers of genes contribute to the heritability for complex diseases. This fact raises fundamental questions about how genetic variation perturbs genetic systems to produce phenotypes. We have proposed one possible model, and it will be important to test this and perhaps others. There are deep challenges to fully understanding the impact of very small effects in organismal systems, so we believe there is great need to develop cell-based model systems that can recapitulate aspects of complex traits. Furthermore, we still have limited understanding of cellular networks, and it will be important to develop highly precise, high-throughput techniques for mapping networks in diverse cell types, especially at the protein level.
We have far better tools than we’ve ever had to do such things, but we’re going to need better ones yet. The advances in biology and chemical biology over the last ten years have made me more optimistic about the state of drug discovery, but that’s on the relative scale. On the absolute scale, we have our work cut out for us. Always have.