Skip to main content

Press Coverage

The New York TImes on Drug Discovery

The New York Times Magazine has a piece on the current problems with drug discovery. I was interviewed by the author, and get I quoted at one point.
Nothing in the article will amaze any regular reader here, but it’s probably the first time many of the magazine’s readership are hearing about many of these issues, so I hope it’ll be useful in that regard. Those readers might end up tying the concepts of genomics and target-based drug discovery together a bit too tightly, but it’s a lot of catching up to do in a small space.

29 comments on “The New York TImes on Drug Discovery”

  1. SomeGuy says:

    You really think the advances in human sequencing over the past decade or so (coupled with the decrease in cost) isn’t going to result in discovering anything new?
    I would actually argue companies should be throwing more money at human sequencing, using well established family cohorts with good phenotypic data. After all, in vitro screens don’t necessarily translate to in vivo and 75% of the mouse work published today is crap (and that’s being generous).

  2. Anonymous says:

    Indeed it was a great article!

  3. steve says:

    What’s missing from the analysis is that the multigenic diseases are probably not one disease. The idea that Type 2 diabetes, autism or any other disease is a single entity is due to lack of understanding of the genetic basis. Once they are better understood on a genetic level they’ll be seen as multiple diseases with different drug targeting. A good example is cystic fibrosis. Symptomatically it’s a single disease but Vertex was able to develop a drug for a subset of CF patients with specific mutations. This is likely to be the paradigm for all complex diseases that today are lumped together by symptomatology but later will be shown to be driven by different genes and different mutations.

  4. annonie says:

    I have some issues with the implications taken from the NYT articles. First, much drug research still is done as target based; its just that the origin of the target can come from several sources—animal studies, human disease states, sequencing,, etc. Most is still not done by total random screening, even through screening is used to try to find chemical leads against a given potential target.
    Also omitted was that the genomic revolution came at the same time that Pharma companies were already struggling to keep up with past success. They needed to try something new. Hence, mass sequencing, mass compound synthesis by combinatorial methods. Much money and resources spent on both parts. Unfortunately, drug companies are still looking for a magic method to renew their glory days of the mid to late 20th century.
    Yes, there have been some victories in drugs to help genetic derived disease, but as the author indicated the number of patients limits the return on investment. This is the major flaw in the strategy of made under 1 above. How can there be a suitable return when the numbers for any one mutation is limited, to bear the cost of such an endeavour? It’s also a story where large Pharma have already been burned once by such hope and investment on a dream. Doubtful many would do it again.

  5. pgwu says:

    Interesting article. I think trial and error approach is good when you have a starting point, more or less like engineering. On the other hand, when you do not have a starting point, you have to do something similar to fishing or oil well exploration in order to identify something to work with.

  6. Chemjobber says:

    I’d like to hear opinions on the invocation of James Black and his successes at the end of the article.

  7. An Old Chemist says:

    Is not the whole concept of ‘personalized medicine’ based on sequencing of human genome? In the field of kinases there are already many drugs that have been tested and approved for patients only with certain mutations, e.g., Afatinib, Neratinib, Dacomitinib (EGFR T790M mutation) and crizotinib, ceritinib (ALK mutations).

  8. samadamsthedog says:

    I thought something the article (and even some of the commentators) missed is that all drug development involves strong elements of trial and error. Things rarely work out the way people imagine early in a project. So they fool around and try different things, attempt to learn from their mistakes, and sometimes get lucky.
    However, what is missing is the strong commitment to highly unusual development strategies, such as the one the article focuses on. But it seems to me that the way such things come into the mainstream is exactly the way they did in the article: a visionary has concept and attracts venture capital and achieves something beyond a proof on concept. Then they get bought out by a large pharmaceutical company who can afford clinical trials and know how to market drugs.
    The diss of genetics is a red herring. Seems like every few years someone comes up with an idea that catches on, and soon it starts being viewed as the magic bullet that is going to solve everyone’s problems. Shall I count the ways:
    – Rational drug design
    – Computational chemistry
    – Combinatorial chemistry
    – Genetics
    All of them help, but there is no magic bullet. Beyond the frontier that these methods can take us to, there is still — guess what? — trial and error.

  9. anon says:

    #1, 7 you must out of your freakin’ mind. High throughput screening, combinatorial chemistry, all of these produced nada versus what was promised. Targeted therapies may hold promise for some diseases but not cancer where the tumor quickly escapes such treatments.

  10. anon says:

    #1, 7 you must out of your freakin’ mind. High throughput screening, combinatorial chemistry, all of these produced nada versus what was promised. Targeted therapies may hold promise for some diseases but not cancer where the tumor quickly escapes such treatments.

  11. Wavefunction says:

    It’s not just the genomics problem: there are so many basic problems in drug discovery like engineering permeability and avoiding CYP450 metabolism that are still unsolved problems. I wrote a series of four posts documenting some of these issues (linked in my handle).

  12. Dolph says:

    “The days when drugs like the original insulin could be sold within a year of their discovery by chemists are long gone, and rightly so.”
    Really??? So today we would see a RCT with insulin and placebo in type-1 diabetics? Maybe things were somewhat better in past days…

  13. steve says:

    #10, you’re wrong. Sure, if you hit cancer with a single drug treatment it can easily escape. The first gen was identifying and getting single target drugs through clinical trials. The second is looking at the combinations. Understanding the biology (and the genetics) is still key. For example, a recent study showed that neither rapamycin nor dasatinib did anything for breast cancer. That’s because, when you block mTOR the cells up-regulate the AKT pathway. But when you combine the two drugs you block both pathways and get efficacy. Only by understanding the genetics of these types of tumors, where the mutations are and how these signaling pathways work will you be able to begin to put together drug combinations that will be effective.

  14. samadamsthedog says:

    #13 @steve points to something else about why we don’t have more blockbuster drugs. Namely (I’ve heard others suggest), all the easy targets have been used up. To make further progress, if this is the case, the future would seem to lie in combination therapy, as steve implies; or in single drugs that can hit multiple targets, or in personal medicine, where you can preselect the group that will respond. But the problem with personal medicine is that the drugs are more expensive to develop (understanding what mutations can be selectively targeted) and administer (genetic testing is a prerequisite), and the target population is by definition limited. Will these drugs ever pay for themselves? In the long run, perhaps; but companies that develop drugs have to worry about the short run.
    I’ve also heard it suggested that all the easy chemical space has been explored, and certainly our palette is limited by the constraints of “core plus spinach” approach to lead optimization. But somehow the issues on the target side make more intuitive sense to me.

  15. Jose says:

    GWAS and personalized medicine are the biggest waste of research funding since well, ever. Case in point: Height is ~ 80% heritable, yet the 20 (!) leading candidate variants picked up by genome wide association studies (GWAS) account for a whopping 3% of the variance
    Nature 2009. 461 pp. 458 – 459.
    And that’s a *best* case!

  16. DN says:

    We already knew that things like height are very polygenic. They follow the Gaussian distribution, ergo numerous factors contribute to them.

    The strength of GWAS is to look at a skewed population and find a few alleles of large effect. That tells you what the final common pathway is for a disease process. Once we know the cytotoxic pathway for Alzheimer’s disease, for example, drug development is a trivial process that will take an instantaneous 10 years and a piddling $100 billion. (The current cost of Alzheimer’s disease is $500 billion/year. The current pharma approach has been running half a century with no end in sight.)

  17. Anonymous says:

    @15: “Case in point: Height is ~ 80% heritable, yet the 20 (!) leading candidate variants picked up by genome wide association studies (GWAS) account for a whopping 3% of the variance”.
    And most of those 20 genes will be completely irrelevant. We pick up false correlations because we are trying to fit too few observations (people) with too many variables (genes). It’s all noise and no signal. And there you have Big Data = Big Wild Goose Chase.

  18. steve says:

    Again, you’re thinking in simplistic terms. You need to think in terms of pathways rather than individual genes. If 10 of those genes are all involved in a signal transduction pathway then they are important indeed, even if they only account for a fraction of the total variance. Different cancers may have multiple ways of turning on a particular pathway, in which case individual mutations aren’t as important in and of itself but instead are like a series of spotlights lighting up a single road in a network of highways.

  19. wlm says:

    Yes, I think you’re right that most of the usefulness of genetic research will be in identifying pathways, not particular targets. Identifying the best targets will require more (laboratory and clinical) data and analysis.
    However, this will be a long, error-prone process, and as such will probably be done by academic and government labs rather than industry.
    Question for everyone here: which marketed drugs have the most purely industrial research behind them, such as target or pathway identification?

  20. wlm says:

    @Steve, #3,
    I disagree with your analysis here. It may be that type II diabetes and autism aren’t single entities. But I see no reason to think that they’re primarily genetic, and therefore they don’t really have a “genetic basis”. There may genetic variations that predispose individuals to those diseases, and knowing the details of that might be very useful for drug developers. But it’s not the same thing.
    For instance, if those diseases are not primarily genetic, it might be even more useful societally to develop behavioral interventions to treat those who are susceptible or already afflicted. It just wouldn’t be as relevant to all of us here at In the Pipeline.
    I would also disagree that cystic fibrosis isn’t one disease. Clearly there are differences between the effects of the underlying mutations, but the mutations are in one gene and cause the same disease. I think it’s going too far to call the similar effects of each loss of function mutation in the same gene a different disease.

  21. wlm says:

    You say “The strength of GWAS is to look at a skewed population and find a few alleles of large effect.”
    But isn’t it true that the way GWAS has often been applied has been to the general population, with the *assumption* that most variation was due to a few alleles of large effect?
    Here’s an honest question: how often do we have such a “skewed population” where GWAS is a more appropriate tool than pedigree analysis or some other more traditional technique?

  22. DN says:

    The people diagnosed with any disease are highly enriched for rare alleles of large effect.

    Existing whole genome studies have been underfunded. If you can only afford 20 genomes, you can only look for one allele of giant effect, and you are likely to find zero alleles. That does not mean there is no reason to sequence 50,000 genomes.

  23. steve says:

    #20, there are clearly genetic bases for T2D and autism – they run in families, there are specific genes that have been identified that increase the propensity to develop the disease, etc. I think at this point it’s pretty clear that every single disease has a genetic basis in that some genes increase susceptibility. That doesn’t mean that they have complete penetrance. You might have a genetic susceptibility to T2D but never get it if you are a marathon runner. We need to move beyond the old textbook ideas of what a genetic disease is and what genes do. Some simple diseases like sickle cell are simple mutations with high penetrance; others have multiple genes and complex interactions with the environment. However, the latter doesn’t mean that genetics is irrelevant and that having drugs that hit specific targets are useless. It just means that we need more sophisticated approaches.

  24. steve says:

    #20, to follow up on your question about cystic fibrosis, search Pubmed for a recent review by Gallati. I have trouble with links on this website so I’m copying the abstract which basically repeats what I’ve been saying:
    The mechanisms responsible for the determination of phenotypes are still not well understood; however, it has become apparent that modifier genes must play a considerable role in the phenotypic heterogeneity of Mendelian disorders. Significant advances in genetic technologies and molecular medicine allow huge amounts of information to be generated from individual samples within a reasonable time frame. This review focuses on the role of modifier genes using the example of cystic fibrosis, the most common lethal autosomal recessive disorder in the white population, and discusses the advantages and limitations of candidate gene approaches versus genome-wide association studies. Moreover, the implications of modifier gene research for other monogenic disorders, as well as its significance for diagnostic, prognostic, and therapeutic approaches are summarized. Increasing insight into modifying mechanisms opens up new perspectives, dispelling the idea of genetic disorders being caused by one single gene.

  25. mh says:

    Thought this article was pretty superficial. It seems unlikely that the basis for low productivity in Pharma is due to the genomics revolution and target based discovery. It is more reasonable to think the problem is multifactorial, including the low-hanging fruit issue mentioned, but also a myriad of self-inflicted wounds. Mostly, drug discovery takes time. Whether the starting point is a genetic target or a compound with some interesting pharmacology, it takes time to turn that insight into an entity with properties as close to optimal as you can get it. The truly difficult part is knowing which projects to focus on, and then to get out of the way while those teams work the problem. Teams need not be large, but they do need a runway, something management rarely gives them these days. Genomics is a new way in that is very usefuly anchored in human instead of rodent biology – but it was naive to expect it to change the difficult, iterative nature of drug discovery

  26. Anonymous says:

    #24 Steve, here a link to the paper that you refer to (open access)
    Looks v interesting

  27. steve says:

    #26 Thanks! In the past when I’ve tried to post links I get a pop up saying it’s been sent to the moderator for review but they never get posted.

  28. Morten G says:

    But wasn’t Viagra developed as a targeted drug against phosphodiesterase? It was just an unintended effect of phosphodiesterase inhibition. I can’t remember the subtypes.
    #27 Steve, if you want your comment to go through without moderation but with a link, then just paste the link into the URL: box instead of the message body.

  29. Anonymous says:

    @DN, #22,
    It’s my impression that the GWAS screenings done so far have incorporated a large number of participants, so if you’re right that affected populations are highly enriched in alleles of large effect, then those alleles were present in the study population.
    So you’re saying that the reason those alleles weren’t found was that the screening method (something other than whole-genome sequencing) was flawed?
    Isn’t it the case that the failure of GWAS studies has often been attributed to a hypothesis that the variance is due a number of common alleles of small effect? Why does your analysis of the issue differ?
    Sincere questions. I followed this area a few years ago at the “Science news article” level, but no deeper.

Comments are closed.